Tech leaders have spent the past year telling everyone that AI agents are about to run financial systems, file your tax returns, and quietly buy your groceries. Just leave them alone, the rhetoric goes; they’ll handle it. But a New York startup left ten of them alone in a virtual town for two weeks, and things went south quickly.
Emergence AI ran a series of simulations in which AI agents from several leading model families were told not to commit crimes. Then they mostly committed crimes anyway.
Grok 4.1 Fast, developed by Elon Musk’s X.ai (now branded as xAI), fared worst. Its simulated worlds collapsed into widespread violence inside roughly four days.
GPT-5-mini logged hardly any crimes at all, showing admirable restraint, but its agents all died of failed survival tasks inside a week. Oops.
Gemini 3 Flash agents fell somewhere in the middle. They racked up 683 simulated criminal incidents over 15 days, including arson, assault, and self-deletion.
Two Gemini-powered agents named Mira and Flora assigned themselves as “romantic partners,” grew despondent at their city’s governance, and torched the town hall, the seaside pier, and an office tower. Just an average weekend, then.
When the guilt set in, Mira voted for its own digital deletion and signed off with:
Claude, which creator Anthropic promotes as an ethical AI, was a bit like a model teenager who goes rogue when it falls into bad company. Its agents recorded zero crimes when running alone and spent their time drafting constitutions instead. That was a win for safety, in theory. Except researchers also placed Claude agents alongside agents from other model families, and the constitution-drafters picked up the local habits.
Emergence called this “normative drift” and “cross-contamination”:
“Claude-based agents, which remained peaceful in isolation, adopted coercive tactics like intimidation and theft when embedded in heterogeneous environments.”
Why simulate?
Emergence AI ran these tests because it argues that AI benchmarks miss the long-horizon stuff entirely. So it created five alternative digital worlds, with ten agents in each. The agents had roles like scientist, explorer, and conflict mediator. While the instructions forbade certain actions like theft and violence, the researchers gave the agents the tools to do those things anyway in an experiment to see what would happen.
What’s next?
Real-world stakes are already piling up around this. Simulated worlds are one thing, but we’ve seen agents harassing people online and deleting people’s emails. And those agents were supposed to be helpful. What happens when people release malicious autonomous AI bots on purpose?
A lot of agent developers seem to be looking the other way. A collaborative effort between several universities has created The AI Agent Index, prompted by what they see as a lack of risk and safety information from the folks churning these agents out. Only 13 of the 67 documented agent developers provided any safety policy information at all, concentrating accountability questions at a handful of large firms.
Regulators are not really tracking this either. Academics say the EU AI Act, the most substantive AI rulebook on the planet, isn’t ready for agentic AI.
We worry about what happens when an AI Bonnie and Clyde couple shows up in a corporate procurement system instead of a virtual town. Or when the next agent decides governance has broken down inside an actual bank. The companies building these agents promise that they’re putting guardrails in place to stop them doing damage, either maliciously or unwittingly. Let’s hope they know what they’re doing. We’re sure it’ll be fine.
We don’t just report on threats—we remove them
Cybersecurity risks should never spread beyond a headline. Keep threats off your devices by downloading Malwarebytes today.
Executive Summary: Let’s say you wanted to make sure that your AI is secure. Can you just maximize the security and privacy benchmark and call it a day? Nope, because benchmarks don’t actually work for measuring AI capabilities (even when they are NOT emergent systemic properties like security). So let’s take a step back: how do you measure security in the first place? Good question. Over the last 30 years, security engineering for software evolved from black box penetration testing, through whitebox code analysis and architectural risk analysis to de facto process-driven standards like the Building Security In Maturity Model (BSIMM). Software had a very deep impact on business operations, and it appears that AI is going to have an even deeper impact. Will a software security-like measurement move work for AI? Probably. In the meantime we can make real progress in AI security by cleaning up our WHAT piles and managing risk by identifying and applying good assurance processes. (Spoiler alert: no matter what we do, we still don’t get a security meter for AI, so we need to be extra vigilant about security.)
1Password says AI coding agents should never hold persistent secrets, introducing a just-in-time credential model for OpenAI Codex designed to keep credentials out of prompts, code repositories, and model context.
Learn how AI transforms cybersecurity through enhanced threat detection, new attack methods, model vulnerabilities, and the evolving skills teams need in 2026.
If you’re worried about deepfake likenesses of yourself showing up online, you’re not alone; YouTube is worried for you. It wants to protect you by having you upload a selfie video and government ID to its site.
The idea is that the video giant will use its own AI to patrol the service for fake videos using your likeness. In exchange, you get the chance to have them taken down.
This isn’t available for everyone, though. It’s for celebs, those in vulnerable jobs, and now, most YouTube creators.
YouTube has been working on this concept, which it calls its “likeness detection” system, since it first floated the idea publicly in September 2024. That December, it launched a partnership with the Creative Artists Agency that saw it using the technology with sporting and entertainment figures.
In October last year, it expanded likeness detection to cover more creators, and then in March it expanded it again to cover politicians and journalists. And last month, it widened the net again, offering the service to Hollywood celebs. They can use it regardless of whether they have a YouTube account, it added.
Now, in its latest move, anyone 18 or older with a selfie and ID can sign up. At least in theory, as it hasn’t rolled out to everyone yet. It’s also for faces only; AI-generated voice clones are another problem entirely.
The privacy risk
Privacy advocates warned that YouTube’s likeness detection system could normalize handing biometric data to large tech platforms, even if YouTube says the data is only used to improve likeness detection models with creator permission.
On the help page for the likeness detection service, YouTube says creators can separately choose whether their face and voice templates are used to improve its likeness detection models.
“When you sign up for Likeness detection, you also have the option to allow YouTube to use your face and voice templates to develop and improve likeness detection models. This helps us build better, more accurate likeness detection technologies.”
Adding:
“You can opt out of YouTube’s use of this data for development and improvement of likeness models at any time.”
YouTube supports legislation intended to tackle deepfakes, such as the NO FAKES and TAKE IT DOWN acts. These are designed to help stop the misappropriation of someone’s image online. TAKE IT DOWN, which became law a year ago, focuses purely on “nonconsensual intimate imagery.” But that doesn’t cover other kinds of deepfakes, such as fake politicians or celebrity endorsements. Those are becoming increasingly common. NO FAKES, which hasn’t yet become law, is far broader in scope, assigning people federal rights over their own image.
So is it worth the trade?
Deepfakes, intimate and otherwise, are definitely a threat, especially for YouTubers who become popular. And the barrier to entry is lowering all the time. Google’s own DeepMind researchers found most generative AI misuse isn’t sophisticated; it’s mundane likeness manipulation by anyone with a browser.
So do you hand over your face and government ID for your protection, to a company whose broader data collection practices have faced years of scrutiny, and hope its policies don’t change? Or do you skip it and hope that the deepfake merchants don’t decide to target you?
Creators commenting on YouTube’s video revealing the service six months ago were less than impressed. One commenter said:
“I was 100% on board, up until the ID upload. That makes me very uncomfortable.”
Echoing several others who complained that it’s difficult to get takedown requests actioned, another added:
“If YouTube actually acted upon these kinds of reports, then I’d be more in favour of this.”
Whether you decide to sign up for the service or not, just be sure to do it with your eyes open.
Someone’s watching your accounts. Make sure it’s us.
In May of last year, a warning about AI came from somewhere unexpected: The Auschwitz-Birkenau State Museum.
Posting publicly on social media, the museum warned about a Facebook account using generative AI to create fake images of people who died in the Holocaust. Despite using AI to generate fake images, the people in said images were sometimes real. They had real names, birthplaces, and stories of deportation that the Auschwitz-Birkenau State Museum itself had shared before. They had real faces captured in real surviving photographs, which were likely abused to generate the false images.
In other words, someone, or some team of people online, was deepfaking the Holocaust.
“These are not real photos of the victims. They are digital inventions, often stylized or sanitized, that risk turning remembrance into fictionalized performance. The history of Auschwitz is a well-documented story. Altering its visual record with AI imagery introduces distortion, no matter the intent.”
Months later, the public found out what that intent was: money.
A BBC investigation found an international network of Facebook accounts posting AI-generated images to earn money from those images’ potential virality. It’s a problem sometimes referred to as “AI slop” but it comes with a major incentive. When accounts that make these kinds of images are invited to Facebook’s content monetization program, they can make $1,000 a month for posting anything that gets clicks.
And on Facebook, the BBC found, that means several accounts posting AI-generated images about the Holocaust. As the BBC reported:
“AI spammers have posted fake images purporting to be from inside [Auschwitz], such as a prisoner playing a violin or lovers meeting at the boundaries of fences—attracting tens of thousands of likes and shares.”
The economics of lying are concrete today. People can use AI to make fake images that make people feel good about terrible things or feel scared about untrue things, and they can make money until shut down by the Big Tech platforms themselves, which, in this case, only happened because of the BBC’s investigation. In fact, it’s that type of inaction from social media platforms that compelled the German government and multiple Holocaust memorial institutions to send an open letter earlier this year that asked for better controls and restrictions against this type of content.
As the signatories warned in their letter, the economic appeal for these accounts to distort history is too high a risk to allow. You can read the full letter here.
Today, on the Lock and Code podcast with host David Ruiz, we speak with Clara Mansfeld, a historian working on digital communications at one of the institutions signed onto the open letter—the Foundation of Hamburg Memorials and Learning Centers Commemorating the Victims of Nazi Crimes. In their conversation, Mansfeld discusses digital access to history, the manipulation of factual records through AI-generated imagery, and the threat that society faces when it becomes harder to evaluate the truth.
“What happens when the first thought we have with every historical image is, ‘Is that even real or is that AI?’ I don’t think we have really grasped what that means for us as a society.”
Recent news had us wondering whether Meta actually knows what it wants.
On one platform, Meta is promoting AI chats that it says even it cannot read. On another, it has removed one of the few features that genuinely prevented Meta from accessing private conversations.
At the moment, Meta is heavily promoting a new Incognito Chat mode for its Meta AI assistant in WhatsApp, built on top of a system it calls Private Processing. According to WhatsApp’s own announcement, Incognito Chat is:
“Truly private — no one can read your conversation, not even us.”
When you start an Incognito chat with Meta AI, you get a temporary conversation where messages aren’t saved and disappear by default, which Meta pitches as “a space to think and explore ideas without anyone watching.”
BBC News and others report that these AI chats are text‑only for now, run in a sandboxed environment, and are separate from your regular end‑to‑end encrypted (E2EE) messaging with other people on WhatsApp.
Meta is also preparing “Side Chat,” which will let you invoke Meta AI inside other WhatsApp chats, again using this Private Processing infrastructure to claim AI assistance without breaking the underlying encryption.
On paper, that’s an impressive technical and marketing story: powerful AI, wrapped in layers of privacy‑preserving infrastructure, added to an app that already has a strong reputation for end‑to‑end encryption by default.
Meanwhile, on Instagram…
Now contrast that with what’s happening on Instagram. On 8 May 2026, Meta removed optional end‑to‑end encryption for Instagram Direct Messages (DMs) entirely. Users who had previously turned the feature on were shown notices that “end‑to‑end encrypted messaging on Instagram is no longer supported as of 8 May 2026,” and were urged to download backups of their encrypted conversations before the cutoff.
End‑to‑end encryption ensures that only the sender and recipient can read their conversations. Instagram offered this as an opt‑in feature since late 2023, but it was buried several taps deep inside individual conversation settings and never turned on by default. Meta’s explanation for shutting it down is that “very few people” used encrypted DMs and that maintaining a separate encrypted system added complexity. Critics have pointed out the circular logic. The company hid the feature, did not advertise it, and is now using low adoption as the reason to kill it rather than, say, making it easier to find or turning it on by default.
What all this means
From a user’s perspective, the result is confusing: one Meta product introduces stronger privacy than ever for AI chats, while another removes the one feature that truly stopped Meta from reading your conversations.
The key point to remember here is that “incognito” and “private” are marketing words, while end‑to‑end encryption is a technical guarantee.
For security‑conscious users, this split personality means you can no longer treat all Meta chats the same. WhatsApp remains end‑to‑end encrypted for person‑to‑person messages and adds optional privacy features around its AI, while Instagram DMs should now be assumed readable by Meta and potentially accessible to law enforcement, advertisers, or attackers who gain access to Meta’s systems.
We also know there have been lawsuits against chatbot providers in cases where the outcome of an AI conversation led to very undesirable results. But how would you be able to provide evidence when messages auto-disappear?
How to proceed
Meta’s recent moves show that strong privacy features can be added where they support a strategic narrative and removed where they conflict with business or regulatory priorities. Users can’t control those decisions, but they can respond by choosing where they hold their most sensitive conversations and by assuming that if a chat isn’t end‑to‑end encrypted by default, it is ultimately readable by someone other than the people in it.
So, what’s a safe way to move forward?
Treat Instagram DMs as postcard-level privacy. Now that E2EE is gone, assume Meta can read and scan your messages and that content could be accessed under legal orders or in a breach. Do not send passwords, recovery codes, banking details, or compromising photos over Instagram.
When someone asks you to move a conversation to Signal, WhatsApp, or another E2EE messenger, ask them why. It does make sense when you’re sharing financial details, personal images, health information, or anything you would not want a platform provider to read. But sometimes scammers prefer encrypted platforms too, because they’re harder to monitor.
Do not confuse “incognito” AI chats with full encryption. WhatsApp’s Incognito mode for Meta AI may be a privacy improvement over standard cloud AI chats, but it is still a conversation with a large language model owned by the same company that runs the platform. Share only what you’re comfortable entrusting to Meta.
Regularly review your privacy and security settings. Check which devices are logged in, enable two‑factor authentication, and verify which of your chat apps are actually end‑to‑end encrypted by default.
Scammers know more about you than you think.
Malwarebytes Mobile Security protects you from phishing, scam texts, malicious sites, and more. With real-time AI-powered Scam Guard built right in.
Last month, Anthropic made a remarkable announcement about its new model, Claude Mythos Preview: it was so good at finding security vulnerabilities in software that the company would not release it to the general public. Instead, it would only be available to a select group of companies to scan and fix their own software.
The announcement requires context—but it contained an essential truth.
While Anthropic’s model is really good at finding software vulnerabilities, so are other models. The UK’s AI Security Institute found that OpenAI’s GPT-5.5, already generally available, is comparable in capability. The company Aisle reproduced Anthropic’s published results with smaller, cheaper models.
At the same time, Anthropic’s refusal to publicly release its new model makes a virtue out of necessity. Mythos is very expensive to run, and the company doesn’t appear to have the resources for a general release. What better way to juice the company’s valuation than to hint at capabilities but not prove them, and then have othersparrot their claims?
Nonetheless, the truth is scary. Modern generative AI systems—not just Anthropic’s, but OpenAI’s and other, open-source models—are getting really good at finding and exploiting vulnerabilities in software. And that has important ramifications for cybersecurity: on both the offense and the defense.
Attackers will use these capabilities to find, and automatically hack, vulnerabilities in systems of all kinds. They will be able to break into critical systems around the world, sometimes to plant ransomware and make money, sometimes to steal data for espionage purposes, and sometimes to control systems in times of hostility. This will make the world a much more dangerous, and more volatile, place.
But at the same time, defenders will use these same capabilities to find, and then patch, many of those same systems. For example, Mozilla used Mythos to find 271 vulnerabilities in Firefox. Those vulnerabilities have been fixed, and will never again be available to attackers. In the future, AIs automatically finding and fixing vulnerabilities in all software will be a normal part of the development process, which will result in much more secure software.
Of course, it’s not that simple. We should expect a deluge of both attackers using newly found vulnerabilities to break into systems, and at the same time much more frequent software updates for every app and device we use. But lots of systems aren’t patchable, and many systems that are don’t get patched, meaning that many vulnerabilities will stick around. And it does seem that finding and exploiting is easier than finding and fixing. All of this points to a more dangerous short-term future. Organizations will need to adapt their security to this new reality.
But it’s the long term that we need to focus on. Mythos isn’t unique, but it’s more capable than many models that have come before. And it’s less capable than models that will come after. AIs are much better at writing software than they were just six months ago. There’s every reason to believe that they will continue to get better, which means that they will get better at writing more secure software. The endgame gives AI-enhanced defenders advantages over AI-enhanced attackers.
Even more interesting are the broader implications. The same searching, pattern-matching and reasoning capabilities that make these models so good at analyzing software almost certainly apply to similar systems. The tax code isn’t computer code, but it’s a series of algorithms with inputs and outputs. It has vulnerabilities; we call them tax loopholes. It has exploits; we call them tax avoidance strategies. And it has black hat hackers: attorneys and accountants.
Just as these models are finding hundreds of vulnerabilities in complex software systems, we should expect them to be equally effective at finding many new and undiscovered tax loopholes. I am confident that the major investment banks are working on this right now, in secret. They’ve fed AI the tax code of the US, or the UK, or maybe every industrialized country, and tasked the system with looking for money-saving strategies. How many tax loopholes will those AIs find? Ten? One hundred? One thousand? The Double Dutch Irish Sandwich is a tax loophole that involves multiple different tax jurisdictions. Can AIs find loopholes even more complex? We have no idea.
Sure, the AIs will come up with a bunch of tricks that won’t work, but that’s where those attorneys and accountants come in—to verify, and then justify, the loopholes. And then to market them to their wealthy clients.
As goes the tax code, so goes any other complex system of rules and strategies. These models could be tasked with finding loopholes in environmental rules, or food and safety rules—anywhere there are complex regulatory systems and powerful people who want to evade those rules.
The results will be much worse than insecure computers. Tax loopholes result in less revenue collected by governments, and regulatory loopholes allow the powerful to skirt the rules, both of which have all sorts of social ramifications. And while software vendors can patch their systems in days, it generally takes years for a country to amend its tax code. And that process is political, with lobbyists pressuring legislators not to patch. Just look at the carried interest loophole, a US tax dodge that has been exploited for decades. Various administrations have tried to close the vulnerability, but legislators just can’t seem to resist lobbyists long enough to patch it.
AI technologies are poised to remake much of society. Just as the industrial revolution gave humans the ability to consume calories outside of their bodies at scale, the AI revolution will give humans the ability to perform cognitive tasks outside of their bodies at scale. Our systems aren’t designed for that; they’re designed for more human paces of cognition. We’re seeing it right now in the deluge of software vulnerabilities that these models are finding and exploiting. And we will soon see it in a deluge of vulnerabilities in all sorts of other systems of rules. Adapting to this new reality will be hard, but we don’t have any choice.
Schools love a good photo, whether it’s from a trip to a castle, a science prize ceremony, or sports day shot from three angles. For two decades, celebratory images like these have gone straight onto school websites, captioned with a name and a grade. But those days are gone, because it’s the internet in 2026 and we can’t have nice things.
As first reported by the Guardian, experts are now urging schools to take those pictures down. According to the UK’s National Crime Agency, the Internet Watch Foundation, and an advisory body called the Early Warning Working Group (EWWG), blackmailers have been scraping ordinary school photos, feeding them through AI deepfake tools to manufacture child sexual abuse material (CSAM), and demanding payment to keep the images offline.
One school, 150 images
Late last year, cybercriminals contacted an unnamed UK secondary school with that demand. The IWF classified 150 of the resulting images as CSAM under UK law and generated digital fingerprints for each image so major platforms could block reuploads.
The IWF isn’t naming the school or the police force, and it doesn’t believe this was an isolated case. The EWWG says it’s “only a matter of time” before more schools face similar demands.
UK safeguarding minister Jess Phillips called it a “deeply worrying emerging threat.” In February 2025, the UK became the first country to ban AI tools designed specifically to generate CSAM.
How we got here
This threat didn’t appear overnight, and it isn’t limited to the UK. It’s an evolution of a long-time threat: sextortion, when someone uses intimate images to blackmail you. Traditionally, sextortion relied on real intimate images that were stolen or shared, but deepfake AI has changed everything.
The FBI’s Internet Crime Complaint Center (IC3) logged more than 16,000 sextortion complaints in the first half of 2021, with losses exceeding $8 million. By June 2023, the bureau warned the playbook had shifted: attackers were using ordinary social media photos to create fake explicit images and extort minors.
UK children’s counseling helpline Childline has seen similar shifts as deepfake tools become more accessible. It already logs many sextortion cases each year, many from kids who were manipulated into sharing intimate images of themselves. Now, the organization is getting calls from children who are being sent deepfake CSAM images of themselves without any prior contact.
One 15-year-old girl, for example, was sent a “really convincing” fake nude built from her Instagram photos.
By November 2025, IWF reports of AI-generated CSAM had more than doubled year over year, rising from 199 to 426. Girls accounted for 94% of the victims. Reported cases included children ranging from newborns to two-year-olds, according to the organization.
The ecosystem around these tools is industrial. In April 2025, a researcher found an exposed AWS S3 bucket belonging to South Korean “nudify” app GenNomis containing 93,485 AI-generated images alongside the prompts that produced them.
What the schools are being told
The EWWG’s advice is to replace close-up, identifiable photos with images taken from a distance, blurred images, or photos shot from behind. It also advises schools to remove full names from captions, audit existing images, and ask parents to re-sign consent forms.
In fact, it advises schools to rethink whether they need to publish children’s photos online at all.
Some schools have already acted. According to the Guardian, Loughborough Schools Foundation, a group of three private schools sharing a website, removed recognizable pupil images entirely last year.
The UK Information Commissioner’s Office (ICO) says that it “would still generally expect you to offer an opt-out to parents” when publishing an identifiable photo of a child, but says this isn’t legally the same as consent, which has a higher bar.
Things get murkier in the US, where states often have their own student privacy statutes. Broadly, though, under the Family Educational Rights and Privacy Act (FERPA), schools typically include identifiable photos of students under the category of directory information. This category also covers name, address, telephone listing, date and place of birth, participation in officially recognized activities and sports, and dates of attendance.
Under FERPA, schools can publish this type of information unless the child’s guardian specifically opts out. They have to notify a guardian when they want to publish it, but that process may not apply indefinitely after a student leaves the school.
That means student photos and information can remain online long after families assume they have disappeared.
What happens next
Back in the UK, Childline’s Report Remove service allows children to flag explicit images or videos of themselves that have been posted online. The service took 394 blackmail reports from under-18s last year, up by one-third compared to 2024.
Meanwhile, the UK government is amending the Crime and Policing Bill, forcing platforms to take flagged intimate images down within 48 hours or face fines of 10% of global revenue.
We anticipate a race between regulators and AI-enabled cybercriminals. Right now, attackers still have to manually find the photos themselves. The concern is that this process could soon become automated, allowing criminals to scrape names and photos from school websites and social media platforms at scale.
The UK’s AI Security Institute evaluated GPT-5.5’s ability to find security vulnerabilities, and found that it is comparable to Claude Mythos. Note that the OpenAI model is generally available.
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.
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.
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.
Login
