The zero-day was designed to bypass 2FA and it was developed by a prominent cybercrime group.

The post Google Detects First AI-Generated Zero-Day Exploit appeared first on SecurityWeek.

The company topped revenue and earnings forecasts for the first quarter of 2026, but its shares plunged more than 20%.

The post Cloudflare Lays Off 1,100 Employees in AI-Driven Restructuring appeared first on SecurityWeek.

Last week on Malwarebytes Labs:

• Microsoft says Edge’s plaintext password behavior is “by design”
• ShinyHunters escalates Canvas attacks with school login defacements
• Massive AI investment scam network spans 15,500 domains
• If a fake moustache can fool age checks, is the Online Safety Act working?
• Google Chrome’s silent 4GB AI download problem
• Attackers adopt JavaScript runtime Bun to spread NWHStealer
• Millions of students’ personal data stolen in major education breach
• Update WhatsApp now: Two new flaws could expose you to malicious files
• Cyberattacks are raising your prices (Lock and Code S07E09)
• Thousands of Facebook accounts stolen by phishing emails sent through Google
• The 2026 World Cup scam economy is already running before the first whistle

Stay safe!

Browse like no one’s watching. 

Malwarebytes Privacy VPN encrypts your connection and never logs what you do, so the next story you read doesn’t have to feel personal. Try it free → 

Last week on Malwarebytes Labs:

• Microsoft says Edge’s plaintext password behavior is “by design”
• ShinyHunters escalates Canvas attacks with school login defacements
• Massive AI investment scam network spans 15,500 domains
• If a fake moustache can fool age checks, is the Online Safety Act working?
• Google Chrome’s silent 4GB AI download problem
• Attackers adopt JavaScript runtime Bun to spread NWHStealer
• Millions of students’ personal data stolen in major education breach
• Update WhatsApp now: Two new flaws could expose you to malicious files
• Cyberattacks are raising your prices (Lock and Code S07E09)
• Thousands of Facebook accounts stolen by phishing emails sent through Google
• The 2026 World Cup scam economy is already running before the first whistle

Stay safe!

Browse like no one’s watching. 

Malwarebytes Privacy VPN encrypts your connection and never logs what you do, so the next story you read doesn’t have to feel personal. Try it free → 

Researchers tracked a large AI‑themed investment scam campaign involving more than 15,000 domains. It uses cloaking and deepfakes to hide from security tools while targeting ordinary users.

Criminals abused the Keitaro ad-tracking platform as part of a cloaking system so real victims see scam content, while security scanners, ad reviewers, and some random visitors see harmless pages, making the operation hard to detect and shut down.

Keitaro is a commercial tracking platform originally meant for digital marketers to manage ad campaigns, test which ads work best, and route visitors to different landing pages.

Because it is feature rich, easy to spin up on regular hosting, and built to filter and route traffic, criminals found they can abuse those capabilities to run scams at scale.

Traffic starts in many places. The scammers used compromised websites, spam emails, social media posts, and online ads, all quietly routing through the same tracking infrastructure.

The scam sites typically promise “Smart AI Trading Technology” or “Intelligent Trading Solutions” and claim consistently high returns, often reinforced with deepfake images or fabricated media to look more credible.

Some parts of the campaign now use deepfake videos and fake interviews with well-known public figures, making it look like a celebrity, or finance expert personally endorses the platform.

Once you follow a link, the cloaking part of the operation kicks in. Cloaking is the trick that makes these scams so hard to see from the outside.

When you click an ad or link, your visit passes through a traffic distribution system (TDS), a kind of router for web visitors that decides which page you see. In these cases, the TDS is connected to the tracker.

The system checks things like:

• Your country/region
• Your device and browser
• Where you came from (Facebook ad, Google ad, email link, etc.)
• Sometimes your IP address reputation or other subtle fingerprints

You’re shown the real investment scam landing page only if you match the “ideal victim” profile (for example, a regular consumer in a target country coming from a social media ad).

Everyone else, like a security researcher, ad platform reviewer, or automated scanner, gets shown a benign page, like a generic blog or placeholder site.

How to stay safe

The best way to stay safe is to stay informed about the tricks scammers use. Learn to spot the red flags that almost always give away scams and phishing emails, and remember:

• There is no such thing as a risk-free, consistently profitable investment. If you’re looking to invest, navigate directly to known, regulated financial institutions.
• Deepfakes are very convincing nowadays, so you will hardly be able to tell the difference between the real celebrity and their deepfake persona.
• Don’t act upon unsolicited investment advice, whether it reaches you by email, social media, or sponsored search results.
• Use an up-to-date, real-time anti-malware solution with a web protection component or a reputable tracking and ad-blocker.
• Don’t act on impulse or under time pressure. Always properly research where your money will be going.

Pro tip: Malwarebytes Scam Guard can help you recognize and analyze scams.

Stop threats before they can do any harm.

Malwarebytes Browser Guard blocks phishing pages and malicious sites automatically. Free, one click to install. Add it to your browser →

Researchers tracked a large AI‑themed investment scam campaign involving more than 15,000 domains. It uses cloaking and deepfakes to hide from security tools while targeting ordinary users.

Criminals abused the Keitaro ad-tracking platform as part of a cloaking system so real victims see scam content, while security scanners, ad reviewers, and some random visitors see harmless pages, making the operation hard to detect and shut down.

Keitaro is a commercial tracking platform originally meant for digital marketers to manage ad campaigns, test which ads work best, and route visitors to different landing pages.

Because it is feature rich, easy to spin up on regular hosting, and built to filter and route traffic, criminals found they can abuse those capabilities to run scams at scale.

Traffic starts in many places. The scammers used compromised websites, spam emails, social media posts, and online ads, all quietly routing through the same tracking infrastructure.

The scam sites typically promise “Smart AI Trading Technology” or “Intelligent Trading Solutions” and claim consistently high returns, often reinforced with deepfake images or fabricated media to look more credible.

Some parts of the campaign now use deepfake videos and fake interviews with well-known public figures, making it look like a celebrity, or finance expert personally endorses the platform.

Once you follow a link, the cloaking part of the operation kicks in. Cloaking is the trick that makes these scams so hard to see from the outside.

When you click an ad or link, your visit passes through a traffic distribution system (TDS), a kind of router for web visitors that decides which page you see. In these cases, the TDS is connected to the tracker.

The system checks things like:

• Your country/region
• Your device and browser
• Where you came from (Facebook ad, Google ad, email link, etc.)
• Sometimes your IP address reputation or other subtle fingerprints

You’re shown the real investment scam landing page only if you match the “ideal victim” profile (for example, a regular consumer in a target country coming from a social media ad).

Everyone else, like a security researcher, ad platform reviewer, or automated scanner, gets shown a benign page, like a generic blog or placeholder site.

How to stay safe

The best way to stay safe is to stay informed about the tricks scammers use. Learn to spot the red flags that almost always give away scams and phishing emails, and remember:

• There is no such thing as a risk-free, consistently profitable investment. If you’re looking to invest, navigate directly to known, regulated financial institutions.
• Deepfakes are very convincing nowadays, so you will hardly be able to tell the difference between the real celebrity and their deepfake persona.
• Don’t act upon unsolicited investment advice, whether it reaches you by email, social media, or sponsored search results.
• Use an up-to-date, real-time anti-malware solution with a web protection component or a reputable tracking and ad-blocker.
• Don’t act on impulse or under time pressure. Always properly research where your money will be going.

Pro tip: Malwarebytes Scam Guard can help you recognize and analyze scams.

Stop threats before they can do any harm.

Malwarebytes Browser Guard blocks phishing pages and malicious sites automatically. Free, one click to install. Add it to your browser →

Today we published our latest Global AI Diffusion Report. The global adoption of artificial intelligence continued to rise in the first quarter of 2026. During the quarter, AI usage increased by 1.5 percentage points from 16.3% to 17.8% of the world’s working age population. Intensity of use among economies with the highest rates of AI diffusion also increased, with 26 economies now exceeding 30% of the working age population using AI.

At the top of Microsoft’s National AI Leaderboard, the UAE continued to lead global AI diffusion at 70.1%. The United States finally started to move up the national rankings, albeit only from 24th to 21st based on a 31.3% usage rate by the working age population.

Notable developments in the quarter included accelerating AI adoption in Asia driven in part by improving AI capabilities in Asian languages. South Korea, Thailand, and Japan saw the greatest movement. More broadly, the quarter brought continued widening of the AI gap between the Global North and South, with usage now at 27.5% in the North and 15.4% in the South. These trends are discussed below, including a deeper dive on the positive impact of enhanced multilingual AI capabilities in Japan.

To track all these trends, we continue to measure AI diffusion as the share of people worldwide between ages 15 and 64 who have used a generative AI product during the reported period. This measure is derived from aggregated and anonymized Microsoft telemetry and adjusted to reflect differences in OS and device-market share, internet penetration, and country population. Additional details on the methodology are available in our AI Diffusion technical paper.^[1]

No single metric is perfect, and this one is no exception. Through the Microsoft AI Economy Institute, we continue to refine how we measure AI diffusion globally, including how adoption varies across countries in ways that best advance priorities such as scientific discovery and productivity gains. For this report, we rely on the strongest cross-country measure available today, and we expect to complement it over time with additional indicators as they emerge and mature.

Sectorally, the quarter saw strengthened AI coding capabilities leading to a dramatic increase in production of software code. This was reflected in production by Anthropic’s Claude Code, the OpenAI’s Codex, and Microsoft’s GitHub Copilot. Git pushes – through which software developers put coding changes online – increased 78% year over year globally. Interestingly, the quarter brought added evidence that, at least for now, AI coding capabilities may be increasing demand for the employment of software developers.

As discussed in more detail in the report, when developer productivity increases, the cost of building software declines. If demand for software is elastic, organizations can respond by building more software across a wider range of use cases. It is still too early to know the full labor-market impact of AI-assisted coding, but the available data shows that in 2025, total U.S. software developer employment reached approximately 2.2 million, rising 8.5% year over year and marking a record high for the profession. Early data for the first quarter of 2026 shows that software developer employment in March 2026 was about 4% higher than in March 2025.

Download the latest Global AI Diffusion report. and explore the data here.

[1] A. Misra, J. Wang, S. McCullers, K. White, and J., L. Ferres, “Measuring AI Diffusion: A Population Normalized Metric for Tracking Global AI Usage,” Nov. 04, 2025, arXiv: arXiv:2511.02781. doi: 10.48550/arXiv.2511.02781. 

The post The state of global AI diffusion in 2026 appeared first on Microsoft On the Issues.

During Q1 2026, the exploit kits leveraged by threat actors to target user systems expanded once again, incorporating new exploits for the Microsoft Office platform, as well as Windows and Linux operating systems.

In this report, we dive into the statistics on published vulnerabilities and exploits, as well as the known vulnerabilities leveraged by popular C2 frameworks throughout Q1 2026.

Statistics on registered vulnerabilities

This section provides statistical data on registered vulnerabilities. The data is sourced from cve.org.

We examine the number of registered CVEs for each month starting from January 2022. The total volume of vulnerabilities continues rising and, according to current reports, the use of AI agents for discovering security issues is expected to further reinforce this upward trend.

Total published vulnerabilities per month from 2022 through 2026 (download)

Next, we analyze the number of new critical vulnerabilities (CVSS > 8.9) over the same period.

Total critical vulnerabilities published per month from 2022 through 2026 (download)

The graph indicates that while the volume of critical vulnerabilities slightly decreased compared to previous years, an upward trend remained clearly visible. At present, we attribute this to the fact that the end of last year was marked by the disclosure of several severe vulnerabilities in web frameworks. The current growth is driven by high-profile issues like React2Shell, the release of exploit frameworks for mobile platforms, and the uncovering of secondary vulnerabilities during the remediation of previously discovered ones. We will be able to test this hypothesis in the next quarter; if correct, the second quarter will show a significant decline, similar to the pattern observed in the previous year.

Exploitation statistics

This section presents statistics on vulnerability exploitation for Q1 2026. The data draws on open sources and our telemetry.

Windows and Linux vulnerability exploitation

In Q1 2026, threat actor toolsets were updated with exploits for new, recently registered vulnerabilities. However, we first examine the list of veteran vulnerabilities that consistently account for the largest share of detections:

• CVE-2018-0802: a remote code execution (RCE) vulnerability in the Equation Editor component
• CVE-2017-11882: another RCE vulnerability also affecting Equation Editor
• CVE-2017-0199: a vulnerability in Microsoft Office and WordPad that allows an attacker to gain control over the system
• CVE-2023-38831: a vulnerability resulting from the improper handling of objects contained within an archive
• CVE-2025-6218: a vulnerability allowing the specification of relative paths to extract files into arbitrary directories, potentially leading to malicious command execution
• CVE-2025-8088: a directory traversal bypass vulnerability during file extraction utilizing NTFS Streams

Among the newcomers, we have observed exploits targeting the Microsoft Office platform and Windows OS components. Notably, these new vulnerabilities exploit logic flaws arising from the interaction between multiple systems, making them technically difficult to isolate within a specific file or library. A list of these vulnerabilities is provided below:

• CVE-2026-21509 and CVE-2026-21514: security feature bypass vulnerabilities: despite Protected View being enabled, a specially crafted file can still execute malicious code without the user’s knowledge. Malicious commands are executed on the victim’s system with the privileges of the user who opened the file.
• CVE-2026-21513: a vulnerability in the Internet Explorer MSHTML engine, which is used to open websites and render HTML markup. The vulnerability involves bypassing rules that restrict the execution of files from untrusted network sources. Interestingly, the data provider for this vulnerability was an LNK file.

These three vulnerabilities were utilized together in a single chain during attacks on Windows-based user systems. While this combination is noteworthy, we believe the widespread use of the entire chain as a unified exploit will likely decline due to its instability. We anticipate that these vulnerabilities will eventually be applied individually as initial entry vectors in phishing campaigns.

Below is the trend of exploit detections on user Windows systems starting from Q1 2025.

Dynamics of the number of Windows users encountering exploits, Q1 2025 – Q1 2026. The number of users who encountered exploits in Q1 2025 is taken as 100% (download)

The vulnerabilities listed here can be leveraged to gain initial access to a vulnerable system and for privilege escalation. This underscores the critical importance of timely software updates.

On Linux devices, exploits for the following vulnerabilities were detected most frequently:

• CVE-2022-0847: a vulnerability known as Dirty Pipe, which enables privilege escalation and the hijacking of running applications
• CVE-2019-13272: a vulnerability caused by improper handling of privilege inheritance, which can be exploited to achieve privilege escalation
• CVE-2021-22555: a heap out-of-bounds write vulnerability in the Netfilter kernel subsystem
• CVE-2023-32233: a vulnerability in the Netfilter subsystem that allows for Use-After-Free conditions and privilege escalation through the improper processing of network requests

Dynamics of the number of Linux users encountering exploits, Q1 2025 – Q1 2026. The number of users who encountered exploits in Q1 2025 is taken as 100% (download)

In the first quarter of 2026, we observed a decrease in the number of detected exploits; however, the detection rates are on the rise relative to the same period last year. For the Linux operating system, the installation of security patches remains critical.

Most common published exploits

The distribution of published exploits by software type in Q1 2026 features an updated set of categories; once again, we see exploits targeting operating systems and Microsoft Office suites.

Distribution of published exploits by platform, Q1 2026 (download)

Vulnerability exploitation in APT attacks

We analyzed which vulnerabilities were utilized in APT attacks during Q1 2026. The ranking provided below includes data based on our telemetry, research, and open sources.

TOP 10 vulnerabilities exploited in APT attacks, Q1 2026 (download)

In Q1 2026, threat actors continued to utilize high-profile vulnerabilities registered in the previous year for APT attacks. The hypothesis we previously proposed has been confirmed: security flaws affecting web applications remain heavily exploited in real-world attacks. However, we are also observing a partial refresh of attacker toolsets. Specifically, during the first quarter of the year, APT campaigns leveraged recently discovered vulnerabilities in Microsoft Office products, edge networking device software, and remote access management systems. Although the most recent vulnerabilities are being exploited most heavily, their general characteristics continue to reinforce established trends regarding the categories of vulnerable software. Consequently, we strongly recommend applying the security patches provided by vendors.

C2 frameworks

In this section, we examine the most popular C2 frameworks used by threat actors and analyze the vulnerabilities targeted by the exploits that interacted with C2 agents in APT attacks.

The chart below shows the frequency of known C2 framework usage in attacks against users during Q1 2026, according to open sources.

TOP 10 C2 frameworks used by APTs to compromise user systems, Q1 2026 (download)

Metasploit has returned to the top of the list of the most common C2 frameworks, displacing Sliver, which now shares the second position with Havoc. These are followed by Covenant and Mythic, the latter of which previously saw greater popularity. After studying open sources and analyzing samples of malicious C2 agents that contained exploits, we determined that the following vulnerabilities were utilized in APT attacks involving the C2 frameworks mentioned above:

• CVE-2023-46604: an insecure deserialization vulnerability allowing for arbitrary code execution within the server process context if the Apache ActiveMQ service is running
• CVE-2024-12356 and CVE-2026-1731: command injection vulnerabilities in BeyondTrust software that allow an attacker to send malicious commands even without system authentication
• CVE-2023-36884: a vulnerability in the Windows Search component that enables command execution on the system, bypassing security mechanisms built into Microsoft Office applications
• CVE-2025-53770: an insecure deserialization vulnerability in Microsoft SharePoint that allows for unauthenticated command execution on the server
• CVE-2025-8088 and CVE-2025-6218: similar directory traversal vulnerabilities that allow files to be extracted from an archive to a predefined path, potentially without the archiving utility displaying any alerts to the user

The nature of the described vulnerabilities indicates that they were exploited to gain initial access to the system. Notably, the majority of these security issues are targeted to bypass authentication mechanisms. This is likely due to the fact that C2 agents are being detected effectively, prompting threat actors to reduce the probability of discovery by utilizing bypass exploits.

Notable vulnerabilities

This section highlights the most significant vulnerabilities published in Q1 2026 that have publicly available descriptions.

CVE-2026-21519: Desktop Window Manager vulnerability

At the core of this vulnerability is a Type Confusion flaw. By attempting to access a resource within the Desktop Window Manager subsystem, an attacker can achieve privilege escalation. A necessary condition for exploiting this issue is existing authorization on the system.

It is worth noting that the DWM subsystem has been under close scrutiny by threat actors for quite some time. Historically, the primary attack vector involves interacting with the NtDComposition* function set.

RegPwn (CVE-2026-21533): a system settings access control vulnerability

CVE-2026-21533 is essentially a logic vulnerability that enables privilege escalation. It stems from the improper handling of privileges within Remote Desktop Services (RDS) components. By modifying service parameters in the registry and replacing the configuration with a custom key, an attacker can elevate privileges to the SYSTEM level. This vulnerability is likely to remain a fixture in threat actor toolsets as a method for establishing persistence and gaining high-level privileges.

CVE-2026-21514: a Microsoft Office vulnerability

This vulnerability was discovered in the wild during attacks on user systems. Notably, an LNK file is used to initiate the exploitation process. CVE-2026-21514 is also a logic issue that allows for bypassing OLE technology restrictions on malicious code execution and the transmission of NetNTLM authentication requests when processing untrusted input.

Clawdbot (CVE-2026-25253): an OpenClaw vulnerability

This vulnerability in the AI agent leaks credentials (authentication tokens) when queried via the WebSocket protocol. It can lead to the compromise of the infrastructure where the agent is installed: researchers have confirmed the ability to access local system data and execute commands with elevated privileges. The danger of CVE-2026-25253 is further compounded by the fact that its exploitation has generated numerous attack scenarios, including the use of prompt injections and ClickFix techniques to install stealers on vulnerable systems.

CVE-2026-34070: LangChain framework vulnerability

LangChain is an open-source framework designed for building applications powered by large language models (LLMs). A directory traversal vulnerability allowed attackers to access arbitrary files within the infrastructure where the framework was deployed. The core of CVE-2026-34070 lies in the fact that certain functions within langchain_core/prompts/loading.py handled configuration files insecurely. This could potentially lead to the processing of files containing malicious data, which could be leveraged to execute commands and expose critical system information or other sensitive files.

CVE-2026-22812: an OpenCode vulnerability

CVE-2026-22812 is another vulnerability identified in AI-assisted coding software. By default, the OpenCode agent provided local access for launching authorized applications via an HTTP server that did not require authentication. Consequently, attackers could execute malicious commands on a vulnerable device with the privileges of the current user.

Conclusion and advice

We observe that the registration of vulnerabilities is steadily gaining momentum in Q1 2026, a trend driven by the widespread development of AI tools designed to identify security flaws across various software types. This trajectory is likely to result not only in a higher volume of registered vulnerabilities but also in an increase in exploit-driven attacks, further reinforcing the critical necessity of timely security patch deployment. Additionally, organizations must prioritize vulnerability management and implement effective defensive technologies to mitigate the risks associated with potential exploitation.

To ensure the rapid detection of threats involving exploit utilization and to prevent their escalation, it is essential to deploy a reliable security solution. Key features of such a tool include continuous infrastructure monitoring, proactive protection, and vulnerability prioritization based on real-world relevance. These mechanisms are integrated into Kaspersky Next, which also provides endpoint security and protection against cyberattacks of any complexity.

During Q1 2026, the exploit kits leveraged by threat actors to target user systems expanded once again, incorporating new exploits for the Microsoft Office platform, as well as Windows and Linux operating systems.

In this report, we dive into the statistics on published vulnerabilities and exploits, as well as the known vulnerabilities leveraged by popular C2 frameworks throughout Q1 2026.

Statistics on registered vulnerabilities

This section provides statistical data on registered vulnerabilities. The data is sourced from cve.org.

We examine the number of registered CVEs for each month starting from January 2022. The total volume of vulnerabilities continues rising and, according to current reports, the use of AI agents for discovering security issues is expected to further reinforce this upward trend.

Total published vulnerabilities per month from 2022 through 2026 (download)

Next, we analyze the number of new critical vulnerabilities (CVSS > 8.9) over the same period.

Total critical vulnerabilities published per month from 2022 through 2026 (download)

The graph indicates that while the volume of critical vulnerabilities slightly decreased compared to previous years, an upward trend remained clearly visible. At present, we attribute this to the fact that the end of last year was marked by the disclosure of several severe vulnerabilities in web frameworks. The current growth is driven by high-profile issues like React2Shell, the release of exploit frameworks for mobile platforms, and the uncovering of secondary vulnerabilities during the remediation of previously discovered ones. We will be able to test this hypothesis in the next quarter; if correct, the second quarter will show a significant decline, similar to the pattern observed in the previous year.

Exploitation statistics

This section presents statistics on vulnerability exploitation for Q1 2026. The data draws on open sources and our telemetry.

Windows and Linux vulnerability exploitation

In Q1 2026, threat actor toolsets were updated with exploits for new, recently registered vulnerabilities. However, we first examine the list of veteran vulnerabilities that consistently account for the largest share of detections:

• CVE-2018-0802: a remote code execution (RCE) vulnerability in the Equation Editor component
• CVE-2017-11882: another RCE vulnerability also affecting Equation Editor
• CVE-2017-0199: a vulnerability in Microsoft Office and WordPad that allows an attacker to gain control over the system
• CVE-2023-38831: a vulnerability resulting from the improper handling of objects contained within an archive
• CVE-2025-6218: a vulnerability allowing the specification of relative paths to extract files into arbitrary directories, potentially leading to malicious command execution
• CVE-2025-8088: a directory traversal bypass vulnerability during file extraction utilizing NTFS Streams

Among the newcomers, we have observed exploits targeting the Microsoft Office platform and Windows OS components. Notably, these new vulnerabilities exploit logic flaws arising from the interaction between multiple systems, making them technically difficult to isolate within a specific file or library. A list of these vulnerabilities is provided below:

• CVE-2026-21509 and CVE-2026-21514: security feature bypass vulnerabilities: despite Protected View being enabled, a specially crafted file can still execute malicious code without the user’s knowledge. Malicious commands are executed on the victim’s system with the privileges of the user who opened the file.
• CVE-2026-21513: a vulnerability in the Internet Explorer MSHTML engine, which is used to open websites and render HTML markup. The vulnerability involves bypassing rules that restrict the execution of files from untrusted network sources. Interestingly, the data provider for this vulnerability was an LNK file.

These three vulnerabilities were utilized together in a single chain during attacks on Windows-based user systems. While this combination is noteworthy, we believe the widespread use of the entire chain as a unified exploit will likely decline due to its instability. We anticipate that these vulnerabilities will eventually be applied individually as initial entry vectors in phishing campaigns.

Below is the trend of exploit detections on user Windows systems starting from Q1 2025.

Dynamics of the number of Windows users encountering exploits, Q1 2025 – Q1 2026. The number of users who encountered exploits in Q1 2025 is taken as 100% (download)

The vulnerabilities listed here can be leveraged to gain initial access to a vulnerable system and for privilege escalation. This underscores the critical importance of timely software updates.

On Linux devices, exploits for the following vulnerabilities were detected most frequently:

• CVE-2022-0847: a vulnerability known as Dirty Pipe, which enables privilege escalation and the hijacking of running applications
• CVE-2019-13272: a vulnerability caused by improper handling of privilege inheritance, which can be exploited to achieve privilege escalation
• CVE-2021-22555: a heap out-of-bounds write vulnerability in the Netfilter kernel subsystem
• CVE-2023-32233: a vulnerability in the Netfilter subsystem that allows for Use-After-Free conditions and privilege escalation through the improper processing of network requests

Dynamics of the number of Linux users encountering exploits, Q1 2025 – Q1 2026. The number of users who encountered exploits in Q1 2025 is taken as 100% (download)

In the first quarter of 2026, we observed a decrease in the number of detected exploits; however, the detection rates are on the rise relative to the same period last year. For the Linux operating system, the installation of security patches remains critical.

Most common published exploits

The distribution of published exploits by software type in Q1 2026 features an updated set of categories; once again, we see exploits targeting operating systems and Microsoft Office suites.

Distribution of published exploits by platform, Q1 2026 (download)

Vulnerability exploitation in APT attacks

We analyzed which vulnerabilities were utilized in APT attacks during Q1 2026. The ranking provided below includes data based on our telemetry, research, and open sources.

TOP 10 vulnerabilities exploited in APT attacks, Q1 2026 (download)

In Q1 2026, threat actors continued to utilize high-profile vulnerabilities registered in the previous year for APT attacks. The hypothesis we previously proposed has been confirmed: security flaws affecting web applications remain heavily exploited in real-world attacks. However, we are also observing a partial refresh of attacker toolsets. Specifically, during the first quarter of the year, APT campaigns leveraged recently discovered vulnerabilities in Microsoft Office products, edge networking device software, and remote access management systems. Although the most recent vulnerabilities are being exploited most heavily, their general characteristics continue to reinforce established trends regarding the categories of vulnerable software. Consequently, we strongly recommend applying the security patches provided by vendors.

C2 frameworks

In this section, we examine the most popular C2 frameworks used by threat actors and analyze the vulnerabilities targeted by the exploits that interacted with C2 agents in APT attacks.

The chart below shows the frequency of known C2 framework usage in attacks against users during Q1 2026, according to open sources.

TOP 10 C2 frameworks used by APTs to compromise user systems, Q1 2026 (download)

Metasploit has returned to the top of the list of the most common C2 frameworks, displacing Sliver, which now shares the second position with Havoc. These are followed by Covenant and Mythic, the latter of which previously saw greater popularity. After studying open sources and analyzing samples of malicious C2 agents that contained exploits, we determined that the following vulnerabilities were utilized in APT attacks involving the C2 frameworks mentioned above:

• CVE-2023-46604: an insecure deserialization vulnerability allowing for arbitrary code execution within the server process context if the Apache ActiveMQ service is running
• CVE-2024-12356 and CVE-2026-1731: command injection vulnerabilities in BeyondTrust software that allow an attacker to send malicious commands even without system authentication
• CVE-2023-36884: a vulnerability in the Windows Search component that enables command execution on the system, bypassing security mechanisms built into Microsoft Office applications
• CVE-2025-53770: an insecure deserialization vulnerability in Microsoft SharePoint that allows for unauthenticated command execution on the server
• CVE-2025-8088 and CVE-2025-6218: similar directory traversal vulnerabilities that allow files to be extracted from an archive to a predefined path, potentially without the archiving utility displaying any alerts to the user

The nature of the described vulnerabilities indicates that they were exploited to gain initial access to the system. Notably, the majority of these security issues are targeted to bypass authentication mechanisms. This is likely due to the fact that C2 agents are being detected effectively, prompting threat actors to reduce the probability of discovery by utilizing bypass exploits.

Notable vulnerabilities

This section highlights the most significant vulnerabilities published in Q1 2026 that have publicly available descriptions.

CVE-2026-21519: Desktop Window Manager vulnerability

At the core of this vulnerability is a Type Confusion flaw. By attempting to access a resource within the Desktop Window Manager subsystem, an attacker can achieve privilege escalation. A necessary condition for exploiting this issue is existing authorization on the system.

It is worth noting that the DWM subsystem has been under close scrutiny by threat actors for quite some time. Historically, the primary attack vector involves interacting with the NtDComposition* function set.

RegPwn (CVE-2026-21533): a system settings access control vulnerability

CVE-2026-21533 is essentially a logic vulnerability that enables privilege escalation. It stems from the improper handling of privileges within Remote Desktop Services (RDS) components. By modifying service parameters in the registry and replacing the configuration with a custom key, an attacker can elevate privileges to the SYSTEM level. This vulnerability is likely to remain a fixture in threat actor toolsets as a method for establishing persistence and gaining high-level privileges.

CVE-2026-21514: a Microsoft Office vulnerability

This vulnerability was discovered in the wild during attacks on user systems. Notably, an LNK file is used to initiate the exploitation process. CVE-2026-21514 is also a logic issue that allows for bypassing OLE technology restrictions on malicious code execution and the transmission of NetNTLM authentication requests when processing untrusted input.

Clawdbot (CVE-2026-25253): an OpenClaw vulnerability

This vulnerability in the AI agent leaks credentials (authentication tokens) when queried via the WebSocket protocol. It can lead to the compromise of the infrastructure where the agent is installed: researchers have confirmed the ability to access local system data and execute commands with elevated privileges. The danger of CVE-2026-25253 is further compounded by the fact that its exploitation has generated numerous attack scenarios, including the use of prompt injections and ClickFix techniques to install stealers on vulnerable systems.

CVE-2026-34070: LangChain framework vulnerability

LangChain is an open-source framework designed for building applications powered by large language models (LLMs). A directory traversal vulnerability allowed attackers to access arbitrary files within the infrastructure where the framework was deployed. The core of CVE-2026-34070 lies in the fact that certain functions within langchain_core/prompts/loading.py handled configuration files insecurely. This could potentially lead to the processing of files containing malicious data, which could be leveraged to execute commands and expose critical system information or other sensitive files.

CVE-2026-22812: an OpenCode vulnerability

CVE-2026-22812 is another vulnerability identified in AI-assisted coding software. By default, the OpenCode agent provided local access for launching authorized applications via an HTTP server that did not require authentication. Consequently, attackers could execute malicious commands on a vulnerable device with the privileges of the current user.

Conclusion and advice

We observe that the registration of vulnerabilities is steadily gaining momentum in Q1 2026, a trend driven by the widespread development of AI tools designed to identify security flaws across various software types. This trajectory is likely to result not only in a higher volume of registered vulnerabilities but also in an increase in exploit-driven attacks, further reinforcing the critical necessity of timely security patch deployment. Additionally, organizations must prioritize vulnerability management and implement effective defensive technologies to mitigate the risks associated with potential exploitation.

To ensure the rapid detection of threats involving exploit utilization and to prevent their escalation, it is essential to deploy a reliable security solution. Key features of such a tool include continuous infrastructure monitoring, proactive protection, and vulnerability prioritization based on real-world relevance. These mechanisms are integrated into Kaspersky Next, which also provides endpoint security and protection against cyberattacks of any complexity.

We have released our latest compliance guide, ISO/IEC 42001:2023 on AWS, which provides practical guidance for organizations designing and operating an Artificial Intelligence Management System (AIMS) using AWS services.

As organizations deploy AI and generative AI workloads in the cloud, aligning with globally recognized standards such as ISO/IEC 42001:2023 becomes an important step toward strengthening AI governance, risk management, and responsible AI practices. This guide helps cloud architects, AI/ML engineers, security teams, compliance leaders, and DevOps practitioners understand how to implement and operate ISO 42001-aligned controls using AWS services while applying the AWS Shared Responsibility Model for AI.

The guide explains how organizations can integrate AWS services into their AIMS to support the requirements defined in ISO 42001:2023 clauses 4–10 and the Annex A control specific to AI systems. It also highlights how AWS AI services, security capabilities, monitoring, and automation can help customers maintain visibility over AI systems, improve operational consistency, and prepare audit-ready evidence.

While AWS provides a secure and compliant cloud infrastructure with built-in responsible AI capabilities, customers remain responsible for defining their AIMS scope, implementing controls, and demonstrating conformity during certification audits.

Inside the guide:

• Overview of the ISO/IEC 42001:2023 framework, including understanding ISO 42001 and its Annexes, and how it relates to the broader ISO AI standards family
• Guidance for integrating with AWS security architecture and applying the AWS Shared Responsibility Model for AI workloads
• Context and scoping considerations for establishing an AIMS on AWS, including defining AI system boundaries within your environment
• Mapping of ISO 42001:2023 clauses 4–10 to AWS services and architectural capabilities, covering organizational context, leadership, planning, support, operation, performance evaluation, and improvement
• Implementation guidance for specific Annex A controls (A.2–A.10), including AI policies, internal organization, resources for AI systems, impact assessments, AI system life cycle management, data governance, transparency for interested parties, use of AI systems, and third-party and customer relationships
• Recommendations for evidence collection, documentation, and audit readiness using AWS native tooling
• Best practices for operationalizing AI compliance activities through automation and infrastructure-as-code

Use this guide to map ISO 42001 clauses and Annex A controls to your AWS environment, automate evidence collection, and reduce the effort involved in preparing for a certification audit.

Download: ISO/IEC 42001:2023 on AWS Compliance Guide

For further assistance, contact AWS Security Assurance Services

If you have feedback about this post, please submit comments in the Comments section below.

Google Chrome has been quietly downloading a 4GB AI model onto users’ devices without asking first.

Security researcher Alexander Hanff, aka ThatPrivacyGuy, reports that Chrome has been silently installing Gemini Nano, Google’s on-device AI model, as a file called weights.bin stored in the OptGuideOnDeviceModel directory within users’ Chrome profiles. This 4GB download happens automatically when Chrome determines your device meets the hardware requirements. It does not ask for consent, and sends no notification—not even one of those annoying cookie banners you’ve learned to dismiss without reading.

The Gemini Nano model powers features like “Help me write” text composition assistance, on-device scam detection, and a Summarizer API that websites can call directly. These features are enabled by default in some recent Chrome versions. And here’s the kicker: if you discover the file and delete it, Chrome simply downloads it again.

Why this matters

Let’s start with the obvious problem: a 4GB download isn’t trivial for everyone. If you’re lucky enough to have unlimited fiber internet, you might not notice. But for users on metered connections, mobile hotspots, or in developing countries where data is expensive, Google just cost them real money without permission. For rural users or those with bandwidth caps, this kind of silent transfer can blow through monthly limits in minutes.

Hanff focuses on the environmental angle. He calculated that if this model were pushed to just 1 billion Chrome users (roughly 30% of Chrome’s user base), the distribution alone would consume 240 gigawatt-hours of energy and generate 60,000 tons of CO2 equivalent. That’s not including actually using the model, just the downloads.

But to us, the most troubling aspect is the broader pattern this represents. Just a few weeks ago, we reported another unsolicited AI invasion on our personal computers discovered by Hanff. He documented how Anthropic’s Claude Desktop app, which silently installed browser integration files across multiple Chromium browsers, including five browsers he didn’t even have installed. The integration would reinstall itself if removed, and it also happened without any meaningful user disclosure.

Hanff argues that both cases likely violate EU privacy law, specifically the ePrivacy Directive’s rules about storing data on user devices and the GDPR’s requirements around transparency and lawful processing. While these claims haven’t been tested in court, they highlight a fundamental tension: can companies just install whatever they want on your computer as long as they say it’s a feature of an app you installed?

Google might argue that having an AI on your device provides better privacy than cloud-based alternatives. Which is generally true, but it does not apply here, since Chrome’s most prominent AI feature—the “AI Mode” pill in the address bar—doesn’t even use the local model. According to Hanff’s analysis, it routes queries to Google’s cloud servers anyway. 

All in all, users see a 4GB local AI model and reasonably assume their data stays private, when in reality, the most visible AI feature sends everything to Google’s servers.

Tech companies need to stop treating silent deployment as acceptable practice. We see no valid excuse for this. Your device is yours. The storage is yours. The bandwidth is yours. And the electricity bill is yours.

What happened to asking for permission? And when I remove it, I want it gone permanently—not automatic reinstallation.

When are the tech giants going to learn that we don’t want to be left discovering after the fact that our devices have become deployment targets for features we never asked for.

Update May 12, 2026 with do it yourself instructions

How to check if the AI model is on your computer (Windows)

1. Open File Explorer
2. At the top of the File Explorer window, click the address bar and paste:

%LOCALAPPDATA%\Google\Chrome\User Data

3. Press Enter
4. Look for a folder named:

OptGuideOnDeviceModel

5. If you see it, Chrome has likely downloaded the AI model

How to check on a Mac

1. Open Finder
2. In the menu bar at the top of the screen, click Go > Go to Folder
3. Paste:

~/Library/Application Support/Google/Chrome/

4. Look for a folder named:

OptGuideOnDeviceModel

Now, remember, this isn’t malware, and its presence doesn’t mean your computer is infected.

Turn off Chrome AI features

This part is relatively easy. You may find online instructions telling you to edit the Windows registry or use Chrome policies, but for most people the simplest and safest approach is to disable the features directly in Chrome.

We don’t recommend manually editing the registry unless you fully understand what you’re doing. Incorrect changes can cause system problems.

Instead, try this first:

1. Open Chrome
2. You can copy and paste this directly into Chrome’s address bar and press Enter:

chrome://settings/ai

3. On the page that opens, you can turn off features such as:
   • “Help me write”
   • AI summaries
   • On-device AI features

 The exact options may vary depending on your Chrome version and region.

4. Then restart Chrome to make sure the changes take effect.

This may stop Chrome from downloading or using the AI model, although some users report the files can return after browser updates.

There is probably no need to delete the files unless you specifically need the storage space.

If chrome://settings/ai does not work, the feature may not yet be available in your region, you may be using a managed work or school account, or your version of Chrome may not support these settings yet.

Do you need to delete the OptGuideOnDeviceModel folder?

You can, but there is probably no need to.

If you disable Chrome’s AI features, the downloaded model should no longer be actively used for those features. Leaving the files in place may also prevent Chrome from downloading them again at a later point.

Browse like no one’s watching. 

Malwarebytes Privacy VPN encrypts your connection and never logs what you do, so the next story you read doesn’t have to feel personal. Try it free → 

Google Chrome has been quietly downloading a 4GB AI model onto users’ devices without asking first.

Security researcher Alexander Hanff, aka ThatPrivacyGuy, reports that Chrome has been silently installing Gemini Nano, Google’s on-device AI model, as a file called weights.bin stored in the OptGuideOnDeviceModel directory within users’ Chrome profiles. This 4GB download happens automatically when Chrome determines your device meets the hardware requirements. It does not ask for consent, and sends no notification—not even one of those annoying cookie banners you’ve learned to dismiss without reading.

The Gemini Nano model powers features like “Help me write” text composition assistance, on-device scam detection, and a Summarizer API that websites can call directly. These features are enabled by default in some recent Chrome versions. And here’s the kicker: if you discover the file and delete it, Chrome simply downloads it again.

Why this matters

Let’s start with the obvious problem: a 4GB download isn’t trivial for everyone. If you’re lucky enough to have unlimited fiber internet, you might not notice. But for users on metered connections, mobile hotspots, or in developing countries where data is expensive, Google just cost them real money without permission. For rural users or those with bandwidth caps, this kind of silent transfer can blow through monthly limits in minutes.

Hanff focuses on the environmental angle. He calculated that if this model were pushed to just 1 billion Chrome users (roughly 30% of Chrome’s user base), the distribution alone would consume 240 gigawatt-hours of energy and generate 60,000 tons of CO2 equivalent. That’s not including actually using the model, just the downloads.

But to us, the most troubling aspect is the broader pattern this represents. Just a few weeks ago, we reported another unsolicited AI invasion on our personal computers discovered by Hanff. He documented how Anthropic’s Claude Desktop app, which silently installed browser integration files across multiple Chromium browsers, including five browsers he didn’t even have installed. The integration would reinstall itself if removed, and it also happened without any meaningful user disclosure.

Hanff argues that both cases likely violate EU privacy law, specifically the ePrivacy Directive’s rules about storing data on user devices and the GDPR’s requirements around transparency and lawful processing. While these claims haven’t been tested in court, they highlight a fundamental tension: can companies just install whatever they want on your computer as long as they say it’s a feature of an app you installed?

Google might argue that having an AI on your device provides better privacy than cloud-based alternatives. Which is generally true, but it does not apply here, since Chrome’s most prominent AI feature—the “AI Mode” pill in the address bar—doesn’t even use the local model. According to Hanff’s analysis, it routes queries to Google’s cloud servers anyway. 

All in all, users see a 4GB local AI model and reasonably assume their data stays private, when in reality, the most visible AI feature sends everything to Google’s servers.

Tech companies need to stop treating silent deployment as acceptable practice. We see no valid excuse for this. Your device is yours. The storage is yours. The bandwidth is yours. And the electricity bill is yours.

What happened to asking for permission? And when I remove it, I want it gone permanently—not automatic reinstallation.

When are the tech giants going to learn that we don’t want to be left discovering after the fact that our devices have become deployment targets for features we never asked for.

Update May 12, 2026 with do it yourself instructions

How to check if the AI model is on your computer (Windows)

1. Open File Explorer
2. At the top of the File Explorer window, click the address bar and paste:

%LOCALAPPDATA%\Google\Chrome\User Data

3. Press Enter
4. Look for a folder named:

OptGuideOnDeviceModel

5. If you see it, Chrome has likely downloaded the AI model

How to check on a Mac

1. Open Finder
2. In the menu bar at the top of the screen, click Go > Go to Folder
3. Paste:

~/Library/Application Support/Google/Chrome/

4. Look for a folder named:

OptGuideOnDeviceModel

Now, remember, this isn’t malware, and its presence doesn’t mean your computer is infected.

Turn off Chrome AI features

This part is relatively easy. You may find online instructions telling you to edit the Windows registry or use Chrome policies, but for most people the simplest and safest approach is to disable the features directly in Chrome.

We don’t recommend manually editing the registry unless you fully understand what you’re doing. Incorrect changes can cause system problems.

Instead, try this first:

1. Open Chrome
2. You can copy and paste this directly into Chrome’s address bar and press Enter:

chrome://settings/ai

3. On the page that opens, you can turn off features such as:
   • “Help me write”
   • AI summaries
   • On-device AI features

 The exact options may vary depending on your Chrome version and region.

4. Then restart Chrome to make sure the changes take effect.

This may stop Chrome from downloading or using the AI model, although some users report the files can return after browser updates.

There is probably no need to delete the files unless you specifically need the storage space.

If chrome://settings/ai does not work, the feature may not yet be available in your region, you may be using a managed work or school account, or your version of Chrome may not support these settings yet.

Do you need to delete the OptGuideOnDeviceModel folder?

You can, but there is probably no need to.

If you disable Chrome’s AI features, the downloaded model should no longer be actively used for those features. Leaving the files in place may also prevent Chrome from downloading them again at a later point.

Browse like no one’s watching. 

Malwarebytes Privacy VPN encrypts your connection and never logs what you do, so the next story you read doesn’t have to feel personal. Try it free → 

Many security leader are asking the same question right now. We already pay for Microsoft Copilot, ChatGPT Enterprise, or Claude. Why buy anything else?

It is a fair question. These are genuinely impressive platforms. And the honest answer is that they can help with some things. Just not the things that matter most for most SOC teams.

This post is a practical guide to where generalist AI earns its place in a SOC and where it runs out of road.

Where generalist AI platforms actually add value

Let’s be direct about what generalist AI platforms do well in a security context.

They are good at drafting, incident summaries, policy documentation, communication templates, and post-mortems. If an analyst needs to translate a technical finding into plain language for an executive, a general-purpose LLM can accelerate that substantially.

They are useful for on-demand research. Asking a question about a CVE, looking up MITRE ATT&CK techniques, or getting a quick primer on an unfamiliar attack class. These are real productivity wins.

They can assist with simple scripting and query construction. Writing a KQL query for a Sentinel rule, generating a Python snippet to parse a log format. Useful, time-saving work.

The common thread is that these are assistance tasks. A human still needs to initiate the process while the AI is a capable co-pilot. And for these use cases, a general-purpose tool is perfectly appropriate.

Where generalist AI runs out of road

The problem is that none of those use cases address the actual constraint facing most SOC teams.

Security teams are not failing because analysts lack knowledge or work too slowly. They are constrained by investigative capacity. Alert volumes are rising. Environments are growing. Attacks are moving faster. And the operating model still assumes humans will triage and investigate the majority of what comes in.

When that assumption breaks down, investigation becomes selective. High-severity alerts get attention. Medium alerts accumulate. Low-severity alerts are deferred or auto-closed. And the uncomfortable truth is that real attacks frequently begin as weak signals. Credential misuse, living-off-the-land techniques, early-stage lateral movement. They rarely present as critical alerts. They appear ordinary until someone actually investigates them.

Generic AI does not fix this. Here is why.

Generalist AI is built for breadth, not depth

ChatGPT and Microsoft Copilot are built for general-purpose text generation. Forensic investigation of a suspicious process execution chain, or a cloud misconfiguration alert at 3am, requires domain-specific knowledge and structured reasoning those platforms were not designed to provide.

Generalist AI assists but does not execute 

Even with a great prompt, a general-purpose AI is accelerating an analyst’s workflow, not replacing the need for one. The investigation still depends on human capacity. And human capacity does not scale as fast as the alert surface grows.

Generalist AI KPIs are increased token usage

Microsoft’s KPI, for example, is token usage. More engagement equals more revenue, regardless of whether your security outcomes improved. That is not a subtle difference. It shapes every product decision, every definition of success. And this can result in very high costs for SOC teams heavily relying on these platforms. This is in stark contrast to Intezer AI SOC which selectively uses LLMs while primarily executing forensic investigations with highly scalable tools and processes. 

Read more about how Intezer Forensic AI SOC follows Anthropic’s best practices.

A practical AI decision framework

Use generalist AI when:

• The task requires drafting or synthesizing text and security context is not critical to the output
• An analyst is researching something unfamiliar and needs a starting point
• The work is advisory and a human will validate and act on every output
• Speed of completion matters more than forensic accuracy

Consider purpose-built AI when:

• You need investigation to happen without an analyst driving every step
• Alert volume has outpaced the team’s capacity to investigate manually
• Medium and low-severity alerts are going uninvestigated because there simply is not time
• You need verdicts accurate enough to act on, not just suggestions to review

The line between these two categories comes down to one question. Do you need AI assistance, or do you need AI execution?

What autonomous execution actually requires

This distinction matters because it shapes what you need from a platform.

Assistance is achievable with a good LLM and a capable prompt. Execution requires something harder: accuracy and forensic depth at investigation time.

General-purpose AI tools and many first-generation AI SOC products rely primarily on LLM analysis and SIEM queries. That is not enough to produce verdicts you can trust without a human checking every one.

Intezer AI SOC is built for the execution side of that line. Automated evidence collection, threat intelligence correlation, network forensics, endpoint forensics, and reverse engineering. That additional depth is what generates the high-confidence verdicts that allow organizations to trust the outcome without a human reviewing every decision.

Below a certain threshold of accuracy and depth, AI assists humans. Above it, organizations can safely offload Tier 1 and Tier 2 work entirely. The threshold is not crossed through breadth. It is crossed through domain specialization and forensic rigor.

Intezer’s investigations produce evidence-based verdicts with 98% accuracy. Up to 2% of alerts are escalated as real incidents while the rest are resolved automatically. That is not a productivity improvement. That is a fundamentally different operating model.

The closed loop of triage and detection engineering

There is one more dimension where general-purpose tools fall short and that is detection engineering.

When a generic AI tool helps an analyst triage an alert, that interaction is largely isolated. The outcome does not feed back into your SIEM rules. It does not surface coverage gaps. It does not help you get better at detecting the same class of threat next time.

Intezer’s investigation outcomes feed directly into detection engineering at the source, continuously identifying broken or noisy rules, flagging coverage gaps against the MITRE ATT&CK framework, and generating deployment-ready detection rules informed by real investigation results. The system improves with every alert it processes. Detection gets better based on evidence, not assumptions.

That closed loop is the difference between a productivity tool and an operating model.

Is a single generalist interface with multiple plugins the answer?

There is also an important architectural point worth making. Generalist AI platforms are increasingly effective at consolidating workflows into a single interface, and in theory, you could extend them into security operations through plugins and MCPs. The building blocks exist.

But in practice, stitching together the specialist capabilities needed for real alert triage such as forensic evidence collection, threat intelligence correlation, reverse engineering, network analysis, etc.  means sourcing, integrating, and maintaining a patchwork of plugins across multiple providers. Each one has its own update cycle, its own failure modes, and its own gaps. The integration burden falls on your team, and keeping it all working reliably over time is its own operational overhead.

At some point the question becomes whether the effort of assembling and maintaining a DIY investigation pipeline inside a generalist platform is worth it — or whether it makes more sense to use a purpose-built system where those capabilities are already unified, tested, and working together out of the box.

The bottom line

Generalist AI platforms have a real role to play in the SOC. Use them for drafting, research, and analyst-driven assistance tasks. It is good at those things and it is likely already paid for.

But do not confuse that with solving the capacity problem. When investigation still depends on human bandwidth, the alert backlog does not disappear. It just accumulates more slowly.

The future SOC is one where AI executes investigation and humans supervise outcomes. Getting there requires technology purpose-built for that job.

Learn more about Intezer AI SOC.

The post Generalist AI for your SOC: When and where to use it appeared first on Intezer.

Today, Microsoft is announcing new agreements with the Center for AI Standards and Innovation (CAISI) in the US and the AI Security Institute (AISI) in the UK to advance the science of AI testing and evaluation, including through collaborative work to test Microsoft’s frontier models, assess safeguards, and help mitigate national security and large-scale public safety risks. These agreements matter because ongoing, rigorous testing is essential to building trust and confidence in advanced AI systems. Well-constructed tests help us understand whether our systems are working as intended and delivering the benefits they are designed to provide. Testing also helps us stay ahead of risks, such as AI-driven cyberattacks and other criminal misuses of AI systems, that can emerge once advanced AI systems are deployed in the world. 

While Microsoft regularly undertakes many types of AI testing on its own, testing for national security and large-scale public safety risks necessarily must be a collaborative endeavor with governments. This type of testing depends on deep technical, scientific, and national security expertise that is uniquely held by institutions like CAISI in the US and AISI in the UK and the government agencies they work with. By combining that government expertise with Microsoft’s experience building and deploying AI systems at global scale, together we are better positioned to anticipate and manage national security and public safety risks in ways that build public trust and confidence in advanced AI systems.  

Improving AI evaluation science through cooperative research and operational experience 

Advancing the science of AI evaluation requires more than isolated research or one-off testing. It depends on sustained collaboration between industry, government, and research institutions. Through our new and expanded partnerships with the US and UK governments—alongside national security–focused evaluations of model capabilities—Microsoft is bringing technical expertise and operational experience to strengthen AI evaluation methods and practical testing foundations.  

• In the US, with CAISI, Microsoft and NIST will collaborate on improving methodologies for adversarial assessments—testing AI systems in ways that probe unexpected behaviors, misuse pathways, and failure modes, much like stress-testing whether airbags, seatbelts, and braking systems work effectively and reliably in safety-critical driving scenarios. This work involves co-developing more systematic and reproducible approaches to evaluation, including shared frameworks, datasets, and workflows for assessing safety, security, and robustness risks in advanced AI systems. It also builds on our AI Red Team’s novel research and tools to detect compromised models at scale. 

• In the UK, with AISI, Microsoft will collaborate on research related to frontier safety and security, including methods for evaluating high-risk capabilities and the effectiveness of the safeguards used to address them. The partnership will also include societal resilience research examining how conversational AI systems interact with users in sensitive contexts.  

These collaborations are designed to improve measurement science, evaluation methodologies, practical testing workflows, and real-world mitigation impact. They reflect a shared commitment to rigorous, practical approaches that can make safeguards stronger and evaluations more reliable. 

Looking ahead 

No organization can address these challenges alone. Our partnerships with CAISI and AISI are a key part of a wider effort to build the institutions, research base, and shared methodologies needed for effective AI testing. This effort also includes: 

• Pursuing research and evaluation in collaboration with other AI institutes globally while helping advance shared priorities and methodologies for testing through the International Network for AI Measurement, Evaluation and Science. 

• Helping deliver industry best practices through the Frontier Model Forum (FMF), an initiative dedicated to advancing the science and practice of frontier AI safety and security. Through the FMF, we are working with other leading AI developers to support independent research, develop shared evaluation methodologies, and promote transparency around risk mitigation strategies.  

• Contributing to MLCommons, a multistakeholder non-profit that develops and operationalizes testing tools such as AILuminate, a family of safety and security benchmarks. In February, we announced efforts underway with institutions in India, Japan, Korea, and Singapore to expand AILuminate to support multilingual, multicultural, and multimodal evaluation, helping to make sure that AI systems work well in the languages and cultural contexts in which people around the world use them. 

As AI capabilities advance, so too must the rigor of the testing and safeguards that underpin them. We will apply what we learn from these partnerships directly into how we design, test, and deploy AI systems, ensuring that progress in evaluation science translates into safer, more secure products for our customers. As these partnerships progress, we will share what we learn and look for opportunities to apply insights and best practices to AI testing more broadly.   

The post Advancing AI evaluation with the Center for AI Standards (US) and Innovation and the AI Security Institute (UK) appeared first on Microsoft On the Issues.

This week on the Lock and Code podcast…

Your prices could be going up because of a little something that one group has started calling the “cyber tax.”

Not a “tax” in any regulatory sense of the word, this newly named “cyber tax” is instead a consequence of the growing number of cyberattacks on small businesses. According to the latest research from the Identity Theft Resource Center, 81% of small- and medium-sized businesses suffered a data breach, a security breach, or both, within the past year. And of those businesses, more than 50% of lost more than $250,000.

According to the most recent data from the US Federal Reserve, the median American family has just $8,000 in savings, meaning that a hit of $250,000 could bankrupt a family and turn their lives upside down. But there’s an interesting layer within this data—the median American family is quite similar to the median American business. In fact, they’re often the exact same person.

The local grocer, the nearby HVAC repair service, the avid cyclist who just opened a bike shop, and the tax professional, and physical therapist helping out neighbors are everyday individuals and family members. They do not have multimillion dollar corporations at their backs, supporting them with legal teams, insurance policies, and dedicated IT support teams.

A loss of $250,000, then, is a potential loss of their business. And to stay afloat, the Identity Theft Resource Center found, for the first time ever, that 38% decided to raise their prices.

“It was near 40% said ‘We actually had to raise prices—we had to pass this cost onto our customers,’” said Eva Velasquez, CEO of the Identity Theft Resource Center. “We’re now really seeing the long-term downstream effects of cyberattacks.”

As frustrating as the cyber tax can be, small businesses themselves are also facing a new wave of cyberattacks, from AI-powered phishing emails so convincing that small business owners can’t tell the legitimate from the illegitimate, to deepfake calls that impersonate the CEO of a three-person company, to supply-chain attacks that target small companies as a way to reach bigger ones.  

Today, on the Lock and Code podcast with host David Ruiz, we speak with Velasquez about cybercrime’s impact on small businesses, the new threats being deployed because of AI, and what is necessary to protect business owners and their consumers.

“Great businesses with great protocols in place can still have a vulnerability exploited because this is what the cyber bad guys are doing all day long. They only have to be right once, whereas small business owners have to be right 100% of the time.”

Tune in today to listen to the full conversation.

Show notes and credits:

Intro Music: “Spellbound” by Kevin MacLeod (incompetech.com)
Licensed under Creative Commons: By Attribution 4.0 License
http://creativecommons.org/licenses/by/4.0/
Outro Music: “Good God” by Wowa (unminus.com)

Listen up—Malwarebytes doesn’t just talk cybersecurity, we provide it.

Protect yourself from online attacks that threaten your identity, your files, your system, and your financial well-being with our exclusive offer for Malwarebytes Premium Security for Lock and Code listeners.

This week on the Lock and Code podcast…

Your prices could be going up because of a little something that one group has started calling the “cyber tax.”

Not a “tax” in any regulatory sense of the word, this newly named “cyber tax” is instead a consequence of the growing number of cyberattacks on small businesses. According to the latest research from the Identity Theft Resource Center, 81% of small- and medium-sized businesses suffered a data breach, a security breach, or both, within the past year. And of those businesses, more than 50% of lost more than $250,000.

According to the most recent data from the US Federal Reserve, the median American family has just $8,000 in savings, meaning that a hit of $250,000 could bankrupt a family and turn their lives upside down. But there’s an interesting layer within this data—the median American family is quite similar to the median American business. In fact, they’re often the exact same person.

The local grocer, the nearby HVAC repair service, the avid cyclist who just opened a bike shop, and the tax professional, and physical therapist helping out neighbors are everyday individuals and family members. They do not have multimillion dollar corporations at their backs, supporting them with legal teams, insurance policies, and dedicated IT support teams.

A loss of $250,000, then, is a potential loss of their business. And to stay afloat, the Identity Theft Resource Center found, for the first time ever, that 38% decided to raise their prices.

“It was near 40% said ‘We actually had to raise prices—we had to pass this cost onto our customers,’” said Eva Velasquez, CEO of the Identity Theft Resource Center. “We’re now really seeing the long-term downstream effects of cyberattacks.”

As frustrating as the cyber tax can be, small businesses themselves are also facing a new wave of cyberattacks, from AI-powered phishing emails so convincing that small business owners can’t tell the legitimate from the illegitimate, to deepfake calls that impersonate the CEO of a three-person company, to supply-chain attacks that target small companies as a way to reach bigger ones.  

Today, on the Lock and Code podcast with host David Ruiz, we speak with Velasquez about cybercrime’s impact on small businesses, the new threats being deployed because of AI, and what is necessary to protect business owners and their consumers.

“Great businesses with great protocols in place can still have a vulnerability exploited because this is what the cyber bad guys are doing all day long. They only have to be right once, whereas small business owners have to be right 100% of the time.”

Tune in today to listen to the full conversation.

Show notes and credits:

Intro Music: “Spellbound” by Kevin MacLeod (incompetech.com)
Licensed under Creative Commons: By Attribution 4.0 License
http://creativecommons.org/licenses/by/4.0/
Outro Music: “Good God” by Wowa (unminus.com)

Listen up—Malwarebytes doesn’t just talk cybersecurity, we provide it.

Protect yourself from online attacks that threaten your identity, your files, your system, and your financial well-being with our exclusive offer for Malwarebytes Premium Security for Lock and Code listeners.

Cybersecurity is at a turning point. Advanced AI models are dramatically accelerating vulnerability discovery and creating conditions ripe for exploitation, underscored by the announcement of Claude Mythos Preview. This marks a shift, and whether this technology will favor defenders or attackers will depend on the choices we make now. 

With the right safeguards, these capabilities can help trusted defenders identify and fix vulnerabilities across critical systems in hospitals, power grids, water, and telecommunications. Released irresponsibly or not properly secured, however, those same capabilities could be abused by malicious actors, threatening the foundations of our digital ecosystem. 

Much of the discussion has rightly focused on risks. As advanced AI models speed up the discovery of vulnerabilities, the way we fix them must speed up too. That means stronger pre-deployment risk assessments and close collaboration between governments, frontier AI developers, software providers, and the broader ecosystem to ensure these tools reduce, rather than increase, cyber risk. This is particularly important as AI systems themselves have become high‑value targets, requiring stronger protection of models, systems, data, and underlying infrastructure. 

This is ultimately an international challenge. Neither software supply chains nor threat actors stop at borders. Neither can our response. Meeting this moment will require shared approaches across countries, sectors, and systems—rooted in trust, shared standards, resilience, and responsible use. 

This moment is also an opportunity. Security has been and remains the top priority at Microsoft. Over the last two years, through our  Secure Future Initiative, we have strengthened our security foundations for this age of AI, in part by using AI to accelerate vulnerability discovery and remediation. We have also invested in fundamental AI for security research, including the development of open-source industry benchmarks that can be used to evaluate whether models are ready for real-world security work. We are accelerating that work through deeper public-private collaboration and in partnership with AI, including Anthropic’s Project Glasswing and OpenAI’s Trusted Access for Cyber program. 

Securing our digital ecosystem with next‑generation AI is within reach but is not automatic.  

Building secure foundations for the era of frontier AI  

Ensuring advanced AI technologies are used to strengthen cybersecurity requires deliberate and urgent action. We are sharing the following recommendations as practical steps governments, industry, and the broader ecosystem can take to ensure these tools, often referred to as “frontier AI”, reinforce the security foundations on which digital societies depend. And we hope to continue to partner with model providers, industry and government so we can work together to improve security outcomes for all. 

1. Reinforce core cybersecurity practices  

Advanced AI can strengthen cybersecurity only when strong, consistent cyber hygiene is already in place. As frontier AI accelerates vulnerability discovery and response, core practices such as rapid patching, access control, and system resilience become more critical, not less. 

Security gains in the frontier AI era depend on close coordination between technology providers advancing new capabilities and the organizations responsible for operating, updating, and securing real‑world systems. Without this interdependence, advanced AI cannot deliver durable improvements in security. No organization can solve these cybersecurity problems alone. 

That is why sustained investment in what we know works remains essential: secure‑by‑design product lifecycles, Zero Trust architectures, multi‑factor authentication, least‑privileged access, and ongoing security training. Broad adoption and harmonization of established cybersecurity frameworks to ensure consistent resilience across AI‑enabled systems. Trusted cloud environments that enable these practices at scale, supporting secure data handling, continuous patching, and the secure deployment of AI‑enabled tools for defenders.  

  2. Release advanced capabilities responsibly  

As frontier AI systems gain reasoning, coding, and agentic capabilities, some of the most serious security risks arise before deployment, including realistic misuse involving multi‑step reasoning, tool use, and reconnaissance. Technical safety benchmarks remain important, but they are insufficient without rigorous, real‑world testing.  

As a result, governments are increasingly establishing pre‑deployment evaluations that combine technical testing with threat modeling. These assessments are most effective when frontier developers work closely with organizations that track national‑security risks. Investing in secure evaluation environments and modern testing methods can help governments keep pace as capabilities advance.  

Responsible release practices, including phased and controlled access, are a critical extension of this approach. Our work with Anthropic in Project Glasswing offers one practical model, enabling trusted defenders to evaluate advanced capabilities in constrained settings prior to broader release. Similarly, OpenAI and Microsoft work closely through Trusted Access for Cyber program, and we already support OpenAI’s use of scoped, early deployments for safety and security testing.  

Responsibility does not end at release. Organizations that deploy frontier models are often best positioned to detect emerging misuse and should monitor, mitigate, and share threat information. Microsoft is working with peers through the Frontier Model Forum to advance best practices for evaluating and managing cyber risk and enable information sharing. Governments should encourage continued industry collaboration to restrict access for identified threat actors and counter adversarial or malicious use of advanced AI. 

  3. Modernize vulnerability management  

AI is changing both the speed of vulnerability discovery and what constitutes meaningful security risk. Faster discovery only improves security if triage, validation, and remediation can keep up. 

As AI accelerates discovery, vulnerability management must shift from tracking raw volume to reducing real‑world risk. That means prioritizing vulnerabilities that are genuinely exploitable, assigning clear responsibility for triage and remediation, and using phased, risk‑based disclosure when private coordination improves safety. Above all, systems must be designed around validation and realistic remediation capacity, not the assumption that more findings automatically lead to better security. 

Developers of frontier AI models should embed vulnerability coordination and disclosure directly into responsible‑release frameworks. And work with governments and industry to ensure findings are routed to the right owners, acted on early, and supported by clear coordination pathways. 

  4. Fix faster: Strengthen and accelerate response and remediation 

As AI accelerates vulnerability discovery, remediation must keep pace. Initiatives such as DARPA’s AI Cyber Challenge show how AI can help both find and fix flaws in open‑source software. Hardening defenses requires investment not just in detection tools but in the people, processes, and infrastructure responsible for fixing vulnerabilities, especially in critical sectors. 

Much of the software underpinning critical infrastructure relies on open‑source components maintained by small teams or volunteers with limited security capacity. A surge in AI‑enabled discovery risks overwhelming existing triage and disclosure processes. Efforts such as the GitHub Secure Open Source Fund, alongside investments by Microsoft and others through the Linux Foundation, Alpha‑Omega, and OpenSSF, are helping maintainers adapt in ways that are practical and aligned with existing workflows.  

Governments should treat remediation capacity as a core resilience priority, including sustained investment in and support for maintainers, surge capacity during large discovery events, and modernized disclosure pathways—recognizing that effective remediation still largely depends on human judgment, coordination, and time.  

  5. Advance AI security internationally 

AI security is essential to deploy AI at scale. Because AI systems, supply chains, and the risks they introduce operate across borders, national approaches alone will not be sufficient. 

Governments and industry should work together to build interoperable international foundations for AI security, including risk evaluation, coordinated vulnerability disclosure, and information sharing. Priorities should include strengthening the defensive use of AI, preventing misuse through shared norms and safeguards, and securing AI systems- and the AI technology stack.  

Global participation is critical. Countries and organizations with limited cybersecurity resources or legacy infrastructure are often the most exposed. International cooperation should prioritize capacity‑building, ensuring that the security benefits of AI are realized broadly and equitably. 

AI security is not just a safeguard; it is an enabler for innovation and growth. By acting collectively and moving quickly, governments and industry can strengthen global digital resilience and unlock the trusted adoption of AI across economies, critical infrastructure, and public services.

Meeting the moment: Use frontier AI capabilities to build trust and confidence  

Meeting this moment is ultimately about trust: not in any single technology or provider, but in our collective ability to introduce advanced AI responsibly.  

Used deliberately and built on strong security foundations, these capabilities can strengthen cybersecurity and reinforce confidence in the systems society depends on. The choice is not between innovation and security but whether we enable them to reinforce one another. 

That outcome is within reach. With governments, industry, and infrastructure operators aligned, advanced AI can be deployed in ways that match real‑world defensive capacity and support trusted, lawful action. Done right and working together, frontier AI can help protect the digital infrastructure that underpins modern life and build lasting confidence in its resilience. 

For more than a decade, the Microsoft Digital Crimes Unit (DCU) has persistently disrupted cybercrime and nation-state threats targeting people, organizations, and critical infrastructure. Explore major disruptions—and the ongoing cases and operations behind them here: Disrupting cyberthreats since 2008 | Microsoft

The post From capability to responsibility: Securing our global digital ecosystem with next‑generation AI appeared first on Microsoft On the Issues.

Unit 42 uncovers high-risk AI browser extensions. Disguised as productivity tools, they steal data, intercept prompts, and exfiltrate passwords. Protect your browser.

The post That AI Extension Helping You Write Emails? It’s Reading Them First appeared first on Unit 42.