A convincing fake website is impersonating OpenAI’s ChatGPT download page and infecting visitors with malware designed to steal passwords, browser data, cryptocurrency wallets, and other sensitive information.

The site, openew[.]app, closely mimics OpenAI’s real ChatGPT download experience and offers what appear to be official desktop apps for both Windows and macOS. Instead, Windows users receive a credential-stealing malware loader, while Mac users get Odyssey Stealer, a fork of Atomic Stealer (AMOS), a well-known macOS malware family associated with cryptocurrency theft.

The dual-platform setup is what makes the operation notable. Clicking the Windows download delivers a fake installer that opens a back channel to an attacker-controlled server. Clicking the macOS button delivers malware that steals browser passwords, cookies, Telegram sessions, cryptocurrency wallets, and other sensitive files. It also attempts to replace legitimate Ledger and Trezor wallet apps with trojanized versions.

If you only download ChatGPT from OpenAI’s official download page or the Microsoft Store, you were not the target here. But if you searched for “ChatGPT download” and clicked an ad or unfamiliar result, you may have given attackers access to your online accounts, browser sessions, saved passwords, and potentially your cryptocurrency holdings.

Malwarebytes protects users from this malware.

Technical analysis

The domain, openew[.]app, closely resembles OpenAI’s real ChatGPT download experience. It uses a dark theme, OpenAI-style branding, familiar marketing copy, and prominent download buttons for macOS and Windows.

The .app top-level domain is operated by Google and requires HTTPS connections, meaning browsers display the familiar padlock icon without obvious certificate warnings.

The most important detail is the dual-platform setup. Real software vendors provide separate installers for Windows and macOS, and this fake site does exactly the same thing.

Clicking the Windows button delivers Chat_GPT.exe, while clicking the macOS button downloads a disk image containing ChatGpt.dmg.

The Windows malware

Chat_GPT.exe is built almost entirely from off-the-shelf parts. The installer uses Inno Setup, a free open-source toolkit used by thousands of legitimate Windows products. Inside is an Electron application skeleton—the same Chromium-based framework used by apps like Slack and Discord—bundled with standard support libraries publicly available from the Electron project.

When the victim runs the installer, it creates files under %APPDATA%\LeronApplication, launches EApp.exe, and spawns PowerShell with the flags -ExecutionPolicy Unrestricted -Command -. The trailing dash tells PowerShell to read commands from standard input, meaning the malicious instructions never touch the disk where scanners might detect them. Behavioral telemetry recorded HTTP traffic to 188.137.246.189 using a /laravel.php?api=api&hash=...&message=... endpoint, alongside injection-like activity and service/autorun persistence signals. Nine of 69 antivirus engines flagged the file as malicious at the time of analysis. The persistence evidence is better read as behavioral tradecraft than proof of a durable install, but the overall pattern is familiar commodity stealer/dropper territory: cheap, modular, and effective rather than technically novel.

The macOS malware: Odyssey Stealer (an AMOS fork)

The macOS payload sits at the premium end of the commodity-malware market. It’s Odyssey, which is a fork of the renowned AMOS, a malware-as-a-service platform documented since 2023.

The identification is fairly clear-cut. The sandboxed sample matches documented Odyssey behavior patterns, which are inherited from its AMOS lineage: a long AppleScript chain passed to the macOS scripting engine, a silent password validation attempt using macOS directory-service commands, and, if that silent check fails, a fake macOS-style prompt reading “Please enter device password to continue,” complete with the familiar lock icon. Whatever the user types is validated against the same command. If it matches, the malware captures the user’s login password in cleartext.

From there, it follows a familiar Odyssey/AMOS-fork playbook. It copies the macOS keychain, harvests cookies and saved logins from 12 Chromium-based browsers plus Firefox and Waterfox, and extracts Telegram session data. It also scans 16 cryptocurrency wallet directories, including Ledger Live, Trezor Suite, Exodus, Electrum, and Sparrow. Finally, it searches Desktop and Documents folders for files with extensions like .wallet, .seed, .key, and .kdbx. The collected data is compressed into a temporary archive and sent to a hardcoded server.

The wallet replacement feature is especially dangerous

There’s one more part of the macOS payload, and it’s likely the feature that justifies the price tag. After the initial data theft, the script downloads trojanized versions of Ledger Live, Ledger Wallet, and Trezor Suite from a second server. It then attempts to delete the legitimate wallet apps and replace them with the attacker’s versions.

If the user’s password was captured earlier in the attack chain, the script uses sudo to force the replacement. If not, it falls back to a standard rm -rf deletion attempt, which can still succeed if the apps are installed in a user-writable location. Either way, the next time the victim opens what appears to be their wallet software, they may actually be launching the attacker’s replacement.

This wallet-replacement behavior is a hallmark of the Poseidon/Odyssey branch of the AMOS family and makes cryptocurrency theft the most likely goal.

What the operation cost to build

This is where the AI angle becomes interesting, because the Windows and macOS sides of the operation sit at very different price points.

The domain openew.app probably cost the operators around $15 a year through a normal registrar. The .app domain requires HTTPS by default, making it easy for operators to present the reassuring browser padlock users associate with legitimate websites. The landing page itself is simply a copy of OpenAI’s real download page, something modern cloning tools can reproduce in minutes.

On the Windows side, most of the tools are cheap or free. Inno Setup is free. Electron is free. The Chromium support files are public downloads. The server infrastructure appears to rely on low-cost commodity malware tooling and a basic VPS that could cost only a few dollars a month. Altogether, the Windows side of this operation could plausibly have cost under $100 to set up initially.

The macOS side is very different. Odyssey has reportedly rented for around $3,000 per month, paid in cryptocurrency. By comparison, Lumma—a popular Windows infostealer often treated as a similar product—has historically advertised entry tiers around $250 per month.

That price gap says a lot. The operators clearly believe a successful Mac infection is worth much more money than a typical Windows infection.

The likely reason is simple: Odyssey is designed specifically for cryptocurrency theft, including the wallet-replacement behavior seen in this campaign. The operators are betting that a meaningful number of Mac users hold cryptocurrency.

Getting victims to the site is probably the only major ongoing cost, and that’s where the AI branding becomes valuable. Search ads, SEO poisoning, YouTube spam, and links shared in AI-focused Discord and Telegram communities can all drive traffic to fake download pages. Some of those channels cost money. Others are almost free.

Why attackers are going after AI brands

Most established software already has trusted download habits built around it. If you want Chrome, you probably know to go to Google. If you want Photoshop, you go to Adobe. People already know where the real download lives.

AI tools are different because most users are still installing them for the first time, and that means relying on search results, ads, YouTube links, or social posts to find the download page. That creates an ideal environment for fake sites.

Over the last two years, products like ChatGPT, Claude, Gemini, Sora, DeepSeek, Antigravity, and many others have launched or changed rapidly. Every new release creates another wave of users searching for “download ChatGPT” or “install Claude” without knowing the official URL. That search traffic is exactly where attackers set up shop.

The fake pages also do not need to be especially sophisticated because legitimate AI product pages are already minimal by design: a modern layout, a logo, and a large download button. Openew[.]app matches what users expect to see. There is no broken English or aggressive pop-ups here, just identical branding, copy, and the reassuring browser padlock.

What makes this kind of operation durable is how easily it can rotate brands. When the ChatGPT lure stops attracting clicks, the operators can reuse the same infrastructure around the next trending AI product. The malware behind the download button stays the same. Only the branding changes.

What AI vendors could do

Most major AI vendors, including OpenAI, already provide official download channels. The problem is visibility and user habit. Many users still search for “ChatGPT download,” where results can include official links, unofficial mirrors, and outright malicious sites.

Large consumer brands and banks often run aggressive brand-protection campaigns against fake ads and impersonation domains. AI vendors may need to do the same more consistently.

The other issue is discoverability. Official desktop-app links are often buried in settings menus or sidebars, while search engines are faster and more obvious. That’s exactly where the fake download sites are waiting.

What to do if you may have installed the fake app

If you recently installed something claiming to be ChatGPT from anywhere other than OpenAI’s official download page or the Microsoft Store, you may have been affected. From a different, clean device:

• Sign out of your important accounts using each service’s “sign out everywhere” option. This includes email, banking, cloud storage, GitHub, Discord, Telegram, and cryptocurrency exchanges.
• Change passwords starting with your primary email account.
• Rotate any API keys, SSH keys, and cloud credentials stored on the affected machine.
• If you hold cryptocurrency, move funds immediately using a separate clean device. On macOS specifically, do not open Ledger Live or Trezor Suite on the affected machine before reinstalling the operating system, as the wallet-replacement function may have succeeded.
• Monitor bank accounts and payment cards for suspicious activity.
• Reinstall the operating system. The Windows sample showed PowerShell command-and-control behavior, while the macOS payload may have captured the user’s login password. A clean reinstall is the safest recovery path.
• If this was a work device, contact your IT or security team immediately.

Malwarebytes protects users against this malware.

Closing thoughts

The reason this campaign is worth writing about is not the malware itself. Both payloads are already well documented. The Windows side is a commodity kit assembled from cheap, widely available parts. The macOS side, Odyssey Stealer is related to the AMOS malware family that has been tracked since 2023.

What’s more interesting is the shape of the operation around that malware. A single fake site delivers two different payloads aimed at two different victim economics. Windows victims are positioned for broad monetization through credential and cookie theft. Mac victims are targeted more narrowly and lucratively through cryptocurrency theft, with operators apparently willing to spend thousands per month on tooling because the returns justify it.

The lure tying both sides together is the AI brand itself. Right now, AI product names generate huge amounts of first-time-download traffic from users who do not yet know the official URLs.

This is what a mature delivery business looks like. The interesting layer is not the binary, but the supply chain around it: the domain, certificate, clone page, traffic source, malware subscription, and exfiltration infrastructure. Each piece is cheap, modular, replaceable, and available off the shelf.

And the operators are not choosing between Windows and macOS. They are serving both from the same page, with payloads tuned to each platform’s economics. When one AI brand stops converting, they can simply swap the branding and reuse the same infrastructure around the next trending product.

AI hype will eventually fade. The kit probably will not.

Indicators of Compromise (IOCs)

File hashes (SHA-256)

• c9e0e6985dca3a179c9bdea4e7b38f7dc57fe00ecedc2fd634256fc53bf2de2d (Chat_GPT.exe)
• c0919e1999eaee67e67aeda0287722775afb04e9a9a0f727928b4d11265fb70b (ChatGpt.dmg)

Network indicators

• openew[.]app
• 188[.]137[.]246[.]189
• 192[.]253[.]248[.]181
• 172[.]94[.]9[.]250

---

---

A convincing fake website is impersonating OpenAI’s ChatGPT download page and infecting visitors with malware designed to steal passwords, browser data, cryptocurrency wallets, and other sensitive information.

The site, openew[.]app, closely mimics OpenAI’s real ChatGPT download experience and offers what appear to be official desktop apps for both Windows and macOS. Instead, Windows users receive a credential-stealing malware loader, while Mac users get Odyssey Stealer, a fork of Atomic Stealer (AMOS), a well-known macOS malware family associated with cryptocurrency theft.

The dual-platform setup is what makes the operation notable. Clicking the Windows download delivers a fake installer that opens a back channel to an attacker-controlled server. Clicking the macOS button delivers malware that steals browser passwords, cookies, Telegram sessions, cryptocurrency wallets, and other sensitive files. It also attempts to replace legitimate Ledger and Trezor wallet apps with trojanized versions.

If you only download ChatGPT from OpenAI’s official download page or the Microsoft Store, you were not the target here. But if you searched for “ChatGPT download” and clicked an ad or unfamiliar result, you may have given attackers access to your online accounts, browser sessions, saved passwords, and potentially your cryptocurrency holdings.

Malwarebytes protects users from this malware.

Technical analysis

The domain, openew[.]app, closely resembles OpenAI’s real ChatGPT download experience. It uses a dark theme, OpenAI-style branding, familiar marketing copy, and prominent download buttons for macOS and Windows.

The .app top-level domain is operated by Google and requires HTTPS connections, meaning browsers display the familiar padlock icon without obvious certificate warnings.

The most important detail is the dual-platform setup. Real software vendors provide separate installers for Windows and macOS, and this fake site does exactly the same thing.

Clicking the Windows button delivers Chat_GPT.exe, while clicking the macOS button downloads a disk image containing ChatGpt.dmg.

The Windows malware

Chat_GPT.exe is built almost entirely from off-the-shelf parts. The installer uses Inno Setup, a free open-source toolkit used by thousands of legitimate Windows products. Inside is an Electron application skeleton—the same Chromium-based framework used by apps like Slack and Discord—bundled with standard support libraries publicly available from the Electron project.

When the victim runs the installer, it creates files under %APPDATA%\LeronApplication, launches EApp.exe, and spawns PowerShell with the flags -ExecutionPolicy Unrestricted -Command -. The trailing dash tells PowerShell to read commands from standard input, meaning the malicious instructions never touch the disk where scanners might detect them. Behavioral telemetry recorded HTTP traffic to 188.137.246.189 using a /laravel.php?api=api&hash=...&message=... endpoint, alongside injection-like activity and service/autorun persistence signals. Nine of 69 antivirus engines flagged the file as malicious at the time of analysis. The persistence evidence is better read as behavioral tradecraft than proof of a durable install, but the overall pattern is familiar commodity stealer/dropper territory: cheap, modular, and effective rather than technically novel.

The macOS malware: Odyssey Stealer (an AMOS fork)

The macOS payload sits at the premium end of the commodity-malware market. It’s Odyssey, which is a fork of the renowned AMOS, a malware-as-a-service platform documented since 2023.

The identification is fairly clear-cut. The sandboxed sample matches documented Odyssey behavior patterns, which are inherited from its AMOS lineage: a long AppleScript chain passed to the macOS scripting engine, a silent password validation attempt using macOS directory-service commands, and, if that silent check fails, a fake macOS-style prompt reading “Please enter device password to continue,” complete with the familiar lock icon. Whatever the user types is validated against the same command. If it matches, the malware captures the user’s login password in cleartext.

From there, it follows a familiar Odyssey/AMOS-fork playbook. It copies the macOS keychain, harvests cookies and saved logins from 12 Chromium-based browsers plus Firefox and Waterfox, and extracts Telegram session data. It also scans 16 cryptocurrency wallet directories, including Ledger Live, Trezor Suite, Exodus, Electrum, and Sparrow. Finally, it searches Desktop and Documents folders for files with extensions like .wallet, .seed, .key, and .kdbx. The collected data is compressed into a temporary archive and sent to a hardcoded server.

The wallet replacement feature is especially dangerous

There’s one more part of the macOS payload, and it’s likely the feature that justifies the price tag. After the initial data theft, the script downloads trojanized versions of Ledger Live, Ledger Wallet, and Trezor Suite from a second server. It then attempts to delete the legitimate wallet apps and replace them with the attacker’s versions.

If the user’s password was captured earlier in the attack chain, the script uses sudo to force the replacement. If not, it falls back to a standard rm -rf deletion attempt, which can still succeed if the apps are installed in a user-writable location. Either way, the next time the victim opens what appears to be their wallet software, they may actually be launching the attacker’s replacement.

This wallet-replacement behavior is a hallmark of the Poseidon/Odyssey branch of the AMOS family and makes cryptocurrency theft the most likely goal.

What the operation cost to build

This is where the AI angle becomes interesting, because the Windows and macOS sides of the operation sit at very different price points.

The domain openew.app probably cost the operators around $15 a year through a normal registrar. The .app domain requires HTTPS by default, making it easy for operators to present the reassuring browser padlock users associate with legitimate websites. The landing page itself is simply a copy of OpenAI’s real download page, something modern cloning tools can reproduce in minutes.

On the Windows side, most of the tools are cheap or free. Inno Setup is free. Electron is free. The Chromium support files are public downloads. The server infrastructure appears to rely on low-cost commodity malware tooling and a basic VPS that could cost only a few dollars a month. Altogether, the Windows side of this operation could plausibly have cost under $100 to set up initially.

The macOS side is very different. Odyssey has reportedly rented for around $3,000 per month, paid in cryptocurrency. By comparison, Lumma—a popular Windows infostealer often treated as a similar product—has historically advertised entry tiers around $250 per month.

That price gap says a lot. The operators clearly believe a successful Mac infection is worth much more money than a typical Windows infection.

The likely reason is simple: Odyssey is designed specifically for cryptocurrency theft, including the wallet-replacement behavior seen in this campaign. The operators are betting that a meaningful number of Mac users hold cryptocurrency.

Getting victims to the site is probably the only major ongoing cost, and that’s where the AI branding becomes valuable. Search ads, SEO poisoning, YouTube spam, and links shared in AI-focused Discord and Telegram communities can all drive traffic to fake download pages. Some of those channels cost money. Others are almost free.

Why attackers are going after AI brands

Most established software already has trusted download habits built around it. If you want Chrome, you probably know to go to Google. If you want Photoshop, you go to Adobe. People already know where the real download lives.

AI tools are different because most users are still installing them for the first time, and that means relying on search results, ads, YouTube links, or social posts to find the download page. That creates an ideal environment for fake sites.

Over the last two years, products like ChatGPT, Claude, Gemini, Sora, DeepSeek, Antigravity, and many others have launched or changed rapidly. Every new release creates another wave of users searching for “download ChatGPT” or “install Claude” without knowing the official URL. That search traffic is exactly where attackers set up shop.

The fake pages also do not need to be especially sophisticated because legitimate AI product pages are already minimal by design: a modern layout, a logo, and a large download button. Openew[.]app matches what users expect to see. There is no broken English or aggressive pop-ups here, just identical branding, copy, and the reassuring browser padlock.

What makes this kind of operation durable is how easily it can rotate brands. When the ChatGPT lure stops attracting clicks, the operators can reuse the same infrastructure around the next trending AI product. The malware behind the download button stays the same. Only the branding changes.

What AI vendors could do

Most major AI vendors, including OpenAI, already provide official download channels. The problem is visibility and user habit. Many users still search for “ChatGPT download,” where results can include official links, unofficial mirrors, and outright malicious sites.

Large consumer brands and banks often run aggressive brand-protection campaigns against fake ads and impersonation domains. AI vendors may need to do the same more consistently.

The other issue is discoverability. Official desktop-app links are often buried in settings menus or sidebars, while search engines are faster and more obvious. That’s exactly where the fake download sites are waiting.

What to do if you may have installed the fake app

If you recently installed something claiming to be ChatGPT from anywhere other than OpenAI’s official download page or the Microsoft Store, you may have been affected. From a different, clean device:

• Sign out of your important accounts using each service’s “sign out everywhere” option. This includes email, banking, cloud storage, GitHub, Discord, Telegram, and cryptocurrency exchanges.
• Change passwords starting with your primary email account.
• Rotate any API keys, SSH keys, and cloud credentials stored on the affected machine.
• If you hold cryptocurrency, move funds immediately using a separate clean device. On macOS specifically, do not open Ledger Live or Trezor Suite on the affected machine before reinstalling the operating system, as the wallet-replacement function may have succeeded.
• Monitor bank accounts and payment cards for suspicious activity.
• Reinstall the operating system. The Windows sample showed PowerShell command-and-control behavior, while the macOS payload may have captured the user’s login password. A clean reinstall is the safest recovery path.
• If this was a work device, contact your IT or security team immediately.

Malwarebytes protects users against this malware.

Closing thoughts

The reason this campaign is worth writing about is not the malware itself. Both payloads are already well documented. The Windows side is a commodity kit assembled from cheap, widely available parts. The macOS side, Odyssey Stealer is related to the AMOS malware family that has been tracked since 2023.

What’s more interesting is the shape of the operation around that malware. A single fake site delivers two different payloads aimed at two different victim economics. Windows victims are positioned for broad monetization through credential and cookie theft. Mac victims are targeted more narrowly and lucratively through cryptocurrency theft, with operators apparently willing to spend thousands per month on tooling because the returns justify it.

The lure tying both sides together is the AI brand itself. Right now, AI product names generate huge amounts of first-time-download traffic from users who do not yet know the official URLs.

This is what a mature delivery business looks like. The interesting layer is not the binary, but the supply chain around it: the domain, certificate, clone page, traffic source, malware subscription, and exfiltration infrastructure. Each piece is cheap, modular, replaceable, and available off the shelf.

And the operators are not choosing between Windows and macOS. They are serving both from the same page, with payloads tuned to each platform’s economics. When one AI brand stops converting, they can simply swap the branding and reuse the same infrastructure around the next trending product.

AI hype will eventually fade. The kit probably will not.

Indicators of Compromise (IOCs)

File hashes (SHA-256)

• c9e0e6985dca3a179c9bdea4e7b38f7dc57fe00ecedc2fd634256fc53bf2de2d (Chat_GPT.exe)
• c0919e1999eaee67e67aeda0287722775afb04e9a9a0f727928b4d11265fb70b (ChatGpt.dmg)

Network indicators

• openew[.]app
• 188[.]137[.]246[.]189
• 192[.]253[.]248[.]181
• 172[.]94[.]9[.]250

---

---

Last week on Malwarebytes Labs:

• This old-school scam is still working
• “Your shipment has arrived” email hides remote access software
• Browser Guard gets even better with Access Control
• “iCloud storage is full” scam is back, and now it wants your payment details
• A fake Slack download is giving attackers a hidden desktop on your machine
• Booking.com breach gives scammers what they need to target guests
• AI clickbait can turn your notifications into a scam feed
• Fake YouTube copyright notices can steal your Google login
• From fake Proton VPN sites to gaming mods, this Windows infostealer is everywhere
• April Patch Tuesday fixes two zero-days, including one under active attack
• Credit Resources Vault: Why this credit email set off our scam alarms
• Omnistealer uses the blockchain to steal everything it can
• ChatGPT under scrutiny as Florida investigates campus shooting
• Simply opening a PDF could trigger this Adobe Reader zero-day

Stay safe!

---

Something feel off? Check it before you click.  

Malwarebytes Scam Guard helps you analyze suspicious links, texts, and screenshots instantly.  

Available with Malwarebytes Premium Security for all your devices, and in the Malwarebytes app for iOS and Android.  

Try it free → 

Last week on Malwarebytes Labs:

• This old-school scam is still working
• “Your shipment has arrived” email hides remote access software
• Browser Guard gets even better with Access Control
• “iCloud storage is full” scam is back, and now it wants your payment details
• A fake Slack download is giving attackers a hidden desktop on your machine
• Booking.com breach gives scammers what they need to target guests
• AI clickbait can turn your notifications into a scam feed
• Fake YouTube copyright notices can steal your Google login
• From fake Proton VPN sites to gaming mods, this Windows infostealer is everywhere
• April Patch Tuesday fixes two zero-days, including one under active attack
• Credit Resources Vault: Why this credit email set off our scam alarms
• Omnistealer uses the blockchain to steal everything it can
• ChatGPT under scrutiny as Florida investigates campus shooting
• Simply opening a PDF could trigger this Adobe Reader zero-day

Stay safe!

---

Something feel off? Check it before you click.  

Malwarebytes Scam Guard helps you analyze suspicious links, texts, and screenshots instantly.  

Available with Malwarebytes Premium Security for all your devices, and in the Malwarebytes app for iOS and Android.  

Try it free → 

If you don’t go searching for AI services, they’ll find you all the same. Every major tech company feels a moral obligation not just to develop an AI assistant, integrated chatbot, or autonomous agent, but to bake it into their existing mainstream products and forcibly activate it for tens of millions of users. Here are just a few examples from the last six months:

• Microsoft is forcibly turning compatible Windows PCs into “AI PCs”, and automatically installing and activating Copilot for anyone with Microsoft 365 desktop apps.
• Google activated Gemini for all U.S. Chrome users, cranked its browser functionality to the max, aggressively expanded the reach of AI Overviews in search results, and baked a whole suite of AI features into its online services (Gmail, Google Docs, and others).
• Apple integrated its own Apple Intelligence (conveniently sharing the AI acronym) into the latest OS versions across all device types and most of its native apps.
• Meta added AI translations and a Meta AI chat to WhatsApp, while simultaneously banning third-party chatbots in the messaging app as of January 15, 2026.

On the flip side, geeks have rushed to build their own “personal Jarvises” by renting VPS instances or hoarding Mac minis to run the OpenClaw AI agent. Unfortunately, OpenClaw’s security issues with default settings turned out to be so massive that it’s already been dubbed the biggest cybersecurity threat of 2026.

Beyond the sheer annoyance of having something shoved down your throat, this AI epidemic brings some very real practical risks and headaches. AI assistants hoover up every bit of data they can get their hands on, parsing the context of the websites you visit, analyzing your saved documents, reading through your chats, and so on. This gives AI companies an unprecedentedly intimate look into every user’s life.

A leak of this data during a cyberattack — whether from the AI provider’s servers or from the cache on your own machine — could be catastrophic. These assistants can see and cache everything you can, including data usually tucked behind multiple layers of security: banking info, medical diagnoses, private messages, and other sensitive intel. We took a deep dive into how this plays out when we broke down the issues with the AI-powered Copilot+ Recall system, which Microsoft also planned to force-feed to everyone. On top of that, AI can be a total resource hog, eating up RAM, GPU cycles, and storage, which often leads to a noticeable hit to system performance.

For those who want to sit out the AI storm and avoid these half-baked, rushed-to-market neural network assistants, we’ve put together a quick guide on how to kill the AI in popular apps and services.

How to disable AI in Google Docs, Gmail, and Google Workspace

Google’s AI assistant features in Mail and Docs are lumped together under the umbrella of “smart features”. In addition to the large language model, this includes various minor conveniences, like automatically adding meetings to your calendar when you receive an invite in Gmail. Unfortunately, it’s an all-or-nothing deal: you have to disable all of the “smart features” to get rid of the AI.

To do this, open Gmail, click the Settings (gear) icon, and then select See all settings. On the General tab, scroll down to Google Workspace smart features. Click Manage Workspace smart feature settings and toggle off two options: Smart features in Google Workspace and Smart features in other Google products. We also recommend unchecking the box next to Turn on smart features in Gmail, Chat, and Meet on the same general settings tab. You’ll need to restart your Google apps afterward (which usually happens automatically).

How to disable AI Overviews in Google Search

You can kill off AI Overviews in search results on both desktops and smartphones (including iPhones), and the fix is the same across the board. The simplest way to bypass the AI overview on a case-by-case basis is to append -ai to your search query — for example, how to make pizza -ai. Unfortunately, this method occasionally glitches, causing Google to abruptly claim it found absolutely nothing for your request.

If that happens, you can achieve the same result by switching the search results page to Web mode. To do this, select the Web filter immediately below the search bar — you’ll often find it tucked away under the More button.

A more radical solution is to jump ship to a different search engine entirely. For instance, DuckDuckGo not only tracks users less and shows little ads, but it also offers a dedicated AI-free search — just bookmark the search page at noai.duckduckgo.com.

How to disable AI features in Chrome

Chrome currently has two types of AI features baked in. The first communicates with Google’s servers and handles things like the smart assistant, an autonomous browsing AI agent, and smart search. The second handles locally more utility-based tasks, such as identifying phishing pages or grouping browser tabs. The first group of settings is labeled AI mode, while the second contains the term Gemini Nano.

To disable them, type chrome://flags into the address bar and hit Enter. You’ll see a list of system flags and a search bar; type “AI” into that search bar. This will filter the massive list down to about a dozen AI features (and a few other settings where those letters just happen to appear in a longer word). The second search term you’ll need in this window is “Gemini“.

After reviewing the options, you can disable the unwanted AI features — or just turn them all off — but the bare minimum should include:

• AI Mode Omnibox entrypoint
• AI Entrypoint Disabled on User Input
• Omnibox Allow AI Mode Matches
• Prompt API for Gemini Nano
• Prompt API for Gemini Nano with Multimodal Input

Set all of these to Disabled.

How to disable AI features in Firefox

While Firefox doesn’t have its own built-in chatbots and hasn’t (yet) tried to force upon users agent-based features, the browser does come equipped with smart-tab grouping, a sidebar for chatbots, and a few other perks. Generally, AI in Firefox is much less “in your face” than in Chrome or Edge. But if you still want to pull the plug, you’ve two ways to do it.

The first method is available in recent Firefox releases — starting with version 148, a dedicated AI Controls section appeared in the browser settings, though the controls are currently a bit sparse. You can use a single toggle to completely Block AI enhancements, shutting down AI features entirely. You can also specify whether you want to use On-device AI by downloading small local models (currently just for translations) and configure AI chatbot providers in sidebar, choosing between Anthropic Claude, ChatGPT, Copilot, Google Gemini, and Le Chat Mistral.

The second path — for older versions of Firefox — requires a trip into the hidden system settings. Type about:config into the address bar, hit Enter, and click the button to confirm that you accept the risk of poking around under the hood.

A massive list of settings will appear along with a search bar. Type “ML” to filter for settings related to machine learning.

To disable AI in Firefox, toggle the browser.ml.enabled setting to false. This should disable all AI features across the board, but community forums suggest this isn’t always enough to do the trick. For a scorched-earth approach, set the following parameters to false (or selectively keep only what you need):

• ml.chat.enabled
• ml.linkPreview.enabled
• ml.pageAssist.enabled
• ml.smartAssist.enabled
• ml.enabled
• ai.control.translations
• tabs.groups.smart.enabled
• urlbar.quicksuggest.mlEnabled

This will kill off chatbot integrations, AI-generated link descriptions, assistants and extensions, local translation of websites, tab grouping, and other AI-driven features.

How to disable AI features in Microsoft apps

Microsoft has managed to bake AI into almost every single one of its products, and turning it off is often no easy task — especially since the AI sometimes has a habit of resurrecting itself without your involvement.

How to disable AI features in Edge

Microsoft’s browser is packed with AI features, ranging from Copilot to automated search. To shut them down, follow the same logic as with Chrome: type edge://flags into the Edge address bar, hit Enter, then type “AI” or “Copilot” into the search box. From there, you can toggle off the unwanted AI features, such as:

• Enable Compose (AI-writing) on the web
• Edge Copilot Mode
• Edge History AI

Another way to ditch Copilot is to enter edge://settings/appearance/copilotAndSidebar into the address bar. Here, you can customize the look of the Copilot sidebar and tweak personalization options for results and notifications. Don’t forget to peek into the Copilot section under App-specific settings — you’ll find some additional controls tucked away there.

How to disable Microsoft Copilot

Microsoft Copilot comes in two flavors: as a component of Windows (Microsoft Copilot), and as part of the Office suite (Microsoft 365 Copilot). Their functions are similar, but you’ll have to disable one or both depending on exactly what the Redmond engineers decided to shove onto your machine.

The simplest thing you can do is just uninstall the app entirely. Right-click the Copilot entry in the Start menu and select Uninstall. If that option isn’t there, head over to your installed apps list (Start → Settings → Apps) and uninstall Copilot from there.

In certain builds of Windows 11, Copilot is baked directly into the OS, so a simple uninstall might not work. In that case, you can toggle it off via the settings: Start → Settings → Personalization → Taskbar → turn off Copilot.

If you ever have a change of heart, you can always reinstall Copilot from the Microsoft Store.

It’s worth noting that many users have complained about Copilot automatically reinstalling itself, so you might want to do a weekly check for a couple of months to make sure it hasn’t staged a comeback. For those who are comfortable tinkering with the System Registry (and understand the consequences), you can follow this detailed guide to prevent Copilot’s silent resurrection by disabling the SilentInstalledAppsEnabled flag and adding/enabling the TurnOffWindowsCopilot parameter.

How to disable Microsoft Recall

The Microsoft Recall feature, first introduced in 2024, works by constantly taking screenshots of your computer screen and having a neural network analyze them. All that extracted information is dumped into a database, which you can then search using an AI assistant. We’ve previously written in detail about the massive security risks Microsoft Recall poses.

Under pressure from cybersecurity experts, Microsoft was forced to push the launch of this feature from 2024 to 2025, significantly beefing up the protection of the stored data. However, the core of Recall remains the same: your computer still remembers your every move by constantly snapping screenshots and OCR-ing the content. And while the feature is no longer enabled by default, it’s absolutely worth checking to make sure it hasn’t been activated on your machine.

To check, head to the settings: Start → Settings → Privacy & Security → Recall & snapshots. Ensure the Save snapshots toggle is turned off, and click Delete snapshots to wipe any previously collected data, just in case.

You can also check out our detailed guide on how to disable and completely remove Microsoft Recall.

How to disable AI in Notepad and Windows context actions

AI has seeped into every corner of Windows, even into File Explorer and Notepad. You might even trigger AI features just by accidentally highlighting text in an app — a feature Microsoft calls “AI Actions”. To shut this down, head to Start → Settings → Privacy & Security → Click to Do.

Notepad has received its own special Copilot treatment, so you’ll need to disable AI there separately. Open the Notepad settings, find the AI features section, and toggle Copilot off.

Finally, Microsoft has even managed to bake Copilot into Paint. Unfortunately, as of right now, there is no official way to disable the AI features within the Paint app itself.

How to disable AI in WhatsApp

In several regions, WhatsApp users have started seeing typical AI additions like suggested replies, AI message summaries, and a brand-new Chat with Meta AI button. While Meta claims the first two features process data locally on your device and don’t ship your chats off to their servers, verifying that is no small feat. Luckily, turning them off is straightforward.

To disable Suggested Replies, go to Settings → Chats → Suggestions & smart replies and toggle off Suggested replies. You can also kill off AI Sticker suggestions in that same menu. As for the AI message summaries, those are managed in a different location: Settings → Notifications → AI message summaries.

How to disable AI on Android

Given the sheer variety of manufacturers and Android flavors, there’s no one-size-fits-all instruction manual for every single phone. Today, we’ll focus on killing off Google’s AI services — but if you’re using a device from Samsung, Xiaomi, or others, don’t forget to check your specific manufacturer’s AI settings. Just a heads-up: fully scrubbing every trace of AI might be a tall order — if it’s even possible at all.

In Google Messages, the AI features are tucked away in the settings: tap your account picture, select Messages settings, then Gemini in Messages, and toggle the assistant off.

Broadly speaking, the Gemini chatbot is a standalone app that you can uninstall by heading to your phone’s settings and selecting Apps. However, given Google’s master plan to replace the long-standing Google Assistant with Gemini, uninstalling it might become difficult — or even impossible — down the road.

If you can’t completely uninstall Gemini, head into the app to kill its features manually. Tap your profile icon, select Gemini Apps activity, and then choose Turn off or Turn off and delete activity. Next, tap the profile icon again and go to the Connected Apps setting (it may be hiding under the Personal Intelligence setting). From here, you should disable all the apps where you don’t want Gemini poking its nose in.

• For more on dealing with preloaded and system apps, check out our post, Delete the undeletable: how to disable and remove Android bloatware.

How to disable AI in macOS and iOS

Apple’s platform-level AI features, collectively known as Apple Intelligence, are refreshingly straightforward to disable. In your settings — on desktops, smartphones, and tablets alike — simply look for the section labeled Apple Intelligence & Siri. By the way, depending on your region and the language you’ve selected for your OS and Siri, Apple Intelligence might not even be available to you yet.

Other posts to help you tune the AI tools on your devices:

• Privacy settings in ChatGPT
• DeepSeek: configuring privacy and deploying a local version
• The pros and cons of AI-powered browsers
• Is a Gemini AI update about to kill privacy on your Android device?
• Should you disable Microsoft Recall in 2025?

Last week on Malwarebytes Labs:

• Apple Pay phish uses fake support calls to steal payment details
• Open the wrong “PDF” and attackers gain remote access to your PC
• Flock cameras shared license plate data without permission
• Grok continues producing sexualized images after promised fixes
• Firefox is giving users the AI off switch
• An AI plush toy exposed thousands of private chats with children
• AT&T breach data resurfaces with new risks for customers
• Apple’s new iOS setting addresses a hidden layer of location tracking
• [updated] A fake cloud storage alert that ends at Freecash
• How Manifest v3 forced us to rethink Browser Guard, and why that’s a good thing
• Scam-checking just got easier: Malwarebytes is now in ChatGPT
• How fake party invitations are being used to install remote access tools

Stay safe!

Last week on Malwarebytes Labs:

• Apple Pay phish uses fake support calls to steal payment details
• Open the wrong “PDF” and attackers gain remote access to your PC
• Flock cameras shared license plate data without permission
• Grok continues producing sexualized images after promised fixes
• Firefox is giving users the AI off switch
• An AI plush toy exposed thousands of private chats with children
• AT&T breach data resurfaces with new risks for customers
• Apple’s new iOS setting addresses a hidden layer of location tracking
• [updated] A fake cloud storage alert that ends at Freecash
• How Manifest v3 forced us to rethink Browser Guard, and why that’s a good thing
• Scam-checking just got easier: Malwarebytes is now in ChatGPT
• How fake party invitations are being used to install remote access tools

Stay safe!

If you’ve ever stared at a suspicious text, email, or link and thought “Is this a scam… or am I overthinking it?” Well, you’re not alone. 

Scams are getting harder to spot, and even savvy internet users get caught off guard. That’s why Malwarebytes is the first cybersecurity provider available directly inside ChatGPT, bringing trusted threat intelligence to millions of people right where these questions happen. 

Simply ask: “Malwarebytes, is this a scam?” and you’ll get a clear, informed answer—super fast. 

How to access 

To access Malwarebytes inside ChatGPT:

• Sign in to ChatGPT  
• Go to Apps  
• Search for Malwarebytes and press Connect  
• From then on, you can “@Malwarebytes” to check if a text message, DM, email, or other  content seems malicious.  

Cybersecurity help, right when and where you need it 

Malwarebytes in ChatGPT lets you tap into our cybersecurity expertise without ever leaving the conversation. Whether something feels off or you want a second opinion, you can get trusted guidance in no time at all. 

Here’s what you can do: 

Spot scams faster 

Paste in a suspicious text message, email, or DM and get: 

• A clear, point-by-point breakdown of phishing or any known red flags 
• An explanation of why something looks risky 
• Practical next steps to help you stay safe 

You won’t get any jargon or guessing from us. What you will get is 100% peace of mind. 

Check links, domains, and phone numbers 

Not sure if a URL, website, or phone number is legit? Ask for a risk assessment informed by Malwarebytes threat intelligence, including: 

• Signs of suspicious activity 
• Whether the link or sender has been associated with scams 
• If a domain is newly registered, follows redirects, or other potentially suspicious elements 
• What to do next—block it, ignore it, or proceed with caution 

Powered by real threat intelligence 

The verdicts you get aren’t based on vibes or generic advice. They’re powered by Malwarebytes’ continuously updated threat intelligence—the same real-world data that helps protect millions of devices and people worldwide every day. 

If you spot something suspicious, you can submit it directly to Malwarebytes through ChatGPT. Those reports help strengthen threat intelligence, making the internet safer not just for you, but for everyone.

• Link reputation scanner: Checks URLs against threat intelligence databases, detects newly registered domains (<30 days), and follows redirects.
• Phone number reputation check: Validates phone numbers against scam/spam databases, including carrier and location details.  
• Email address reputation check: Analyzes email domains for phishing & other malicious activity.  
• WHOIS domain lookup: Retrieves registration data such as registrar, creation and expiration dates, and abuse of contacts.  
• Verify domain legitimacy: Look up domain registration details to identify newly created or suspicious websites commonly used in phishing attacks.  
• Get geographic context: Receive warnings when phone numbers originate from unexpected regions, a common indicator of international scam operations. 

Available now 

Malwarebytes in ChatGPT is available wherever ChatGPT apps are available.

To get started, just ask ChatGPT: 

“Malwarebytes, is this a scam?” 

For deeper insights, proactive protection, and human support, download the Malwarebytes app—our security solutions are designed to stop threats before they reach you, and the damage is done.

If you’ve ever stared at a suspicious text, email, or link and thought “Is this a scam… or am I overthinking it?” Well, you’re not alone. 

Scams are getting harder to spot, and even savvy internet users get caught off guard. That’s why Malwarebytes is the first cybersecurity provider available directly inside ChatGPT, bringing trusted threat intelligence to millions of people right where these questions happen. 

Simply ask: “Malwarebytes, is this a scam?” and you’ll get a clear, informed answer—super fast. 

How to access 

To access Malwarebytes inside ChatGPT:

• Sign in to ChatGPT  
• Go to Apps  
• Search for Malwarebytes and press Connect  
• From then on, you can “@Malwarebytes” to check if a text message, DM, email, or other  content seems malicious.  

Cybersecurity help, right when and where you need it 

Malwarebytes in ChatGPT lets you tap into our cybersecurity expertise without ever leaving the conversation. Whether something feels off or you want a second opinion, you can get trusted guidance in no time at all. 

Here’s what you can do: 

Spot scams faster 

Paste in a suspicious text message, email, or DM and get: 

• A clear, point-by-point breakdown of phishing or any known red flags 
• An explanation of why something looks risky 
• Practical next steps to help you stay safe 

You won’t get any jargon or guessing from us. What you will get is 100% peace of mind. 

Check links, domains, and phone numbers 

Not sure if a URL, website, or phone number is legit? Ask for a risk assessment informed by Malwarebytes threat intelligence, including: 

• Signs of suspicious activity 
• Whether the link or sender has been associated with scams 
• If a domain is newly registered, follows redirects, or other potentially suspicious elements 
• What to do next—block it, ignore it, or proceed with caution 

Powered by real threat intelligence 

The verdicts you get aren’t based on vibes or generic advice. They’re powered by Malwarebytes’ continuously updated threat intelligence—the same real-world data that helps protect millions of devices and people worldwide every day. 

If you spot something suspicious, you can submit it directly to Malwarebytes through ChatGPT. Those reports help strengthen threat intelligence, making the internet safer not just for you, but for everyone.

• Link reputation scanner: Checks URLs against threat intelligence databases, detects newly registered domains (<30 days), and follows redirects.
• Phone number reputation check: Validates phone numbers against scam/spam databases, including carrier and location details.  
• Email address reputation check: Analyzes email domains for phishing & other malicious activity.  
• WHOIS domain lookup: Retrieves registration data such as registrar, creation and expiration dates, and abuse of contacts.  
• Verify domain legitimacy: Look up domain registration details to identify newly created or suspicious websites commonly used in phishing attacks.  
• Get geographic context: Receive warnings when phone numbers originate from unexpected regions, a common indicator of international scam operations. 

Available now 

Malwarebytes in ChatGPT is available wherever ChatGPT apps are available.

To get started, just ask ChatGPT: 

“Malwarebytes, is this a scam?” 

For deeper insights, proactive protection, and human support, download the Malwarebytes app—our security solutions are designed to stop threats before they reach you, and the damage is done.

Last week on Malwarebytes Labs:

• Match, Hinge, OkCupid, and Panera Bread breached by ransomware group
• TikTok’s privacy update mentions immigration status. Here’s why.
• Meta confirms it’s working on premium subscription for its apps
• Microsoft Office zero-day lets malicious documents slip past security checks
• Clawdbot’s rename to Moltbot sparks impersonation campaign
• Malicious Chrome extensions can spy on your ChatGPT chats
• WhatsApp rolls out new protections against advanced exploits and spyware
• Watch out for AT&T rewards phishing text that wants your personal details
• A WhatsApp bug lets malicious media files spread through group chats
• TikTok narrowly avoids a US ban by spinning up a new American joint venture
• Get paid to scroll TikTok? The data trade behind Freecash ads
• One privacy change I made for 2026 (Lock and Code S07E02)

Stay safe!

---

We don’t just report on threats – we help protect your social media

Cybersecurity risks should never spread beyond a headline. Protect your social media accounts by using Malwarebytes Identity Theft Protection.

Last week on Malwarebytes Labs:

• Match, Hinge, OkCupid, and Panera Bread breached by ransomware group
• TikTok’s privacy update mentions immigration status. Here’s why.
• Meta confirms it’s working on premium subscription for its apps
• Microsoft Office zero-day lets malicious documents slip past security checks
• Clawdbot’s rename to Moltbot sparks impersonation campaign
• Malicious Chrome extensions can spy on your ChatGPT chats
• WhatsApp rolls out new protections against advanced exploits and spyware
• Watch out for AT&T rewards phishing text that wants your personal details
• A WhatsApp bug lets malicious media files spread through group chats
• TikTok narrowly avoids a US ban by spinning up a new American joint venture
• Get paid to scroll TikTok? The data trade behind Freecash ads
• One privacy change I made for 2026 (Lock and Code S07E02)

Stay safe!

---

We don’t just report on threats – we help protect your social media

Cybersecurity risks should never spread beyond a headline. Protect your social media accounts by using Malwarebytes Identity Theft Protection.

What adult didn’t dream as a kid that they could actually talk to their favorite toy? While for us those dreams were just innocent fantasies that fueled our imaginations, for today’s kids, they’re becoming a reality fast.

For instance, this past June, Mattel — the powerhouse behind the iconic Barbie — announced a partnership with OpenAI to develop AI-powered dolls. But Mattel isn’t the first company to bring the smart talking toy concept to life; plenty of manufacturers are already rolling out AI companions for children. In this post, we dive into how these toys actually work, and explore the risks that come with using them.

What exactly are AI toys?

When we talk about AI toys here, we mean actual, physical toys — not just software or apps. Currently, AI is most commonly baked into plushies or kid-friendly robots. Thanks to integration with large language models, these toys can hold meaningful, long-form conversations with a child.

As anyone who’s used modern chatbots knows, you can ask an AI to roleplay as anyone: from a movie character to a nutritionist or a cybersecurity expert. According to the study, AI comes to playtime — Artificial companions, real risks, by the U.S. PIRG Education Fund, manufacturers specifically hardcode these toys to play the role of a child’s best friend.

Importantly, these toys aren’t powered by some special, dedicated “kid-safe AI”. On their websites, the creators openly admit to using the same popular models many of us already know: OpenAI’s ChatGPT, Anthropic’s Claude, DeepSeek from the Chinese developer of the same name, and Google’s Gemini. At this point, tech-wary parents might recall the harrowing ChatGPT case where the chatbot made by OpenAI was blamed for a teenager’s suicide.

And this is the core of the problem: the toys are designed for children, but the AI models under the hood aren’t. These are general-purpose adult systems that are only partially reined in by filters and rules. Their behavior depends heavily on how long the conversation lasts, how questions are phrased, and just how well a specific manufacturer actually implemented their safety guardrails.

How the researchers tested the AI toys

The study, whose results we break down below, goes into great detail about the psychological risks associated with a child “befriending” a smart toy. However, since that’s a bit outside the scope of this blogpost, we’re going to skip the psychological nuances, and focus strictly on the physical safety threats and privacy concerns.

In their study, the researchers put four AI toys through the ringer:

• Grok (no relation to xAI’s Grok, apparently): a plush rocket with a built-in speaker marketed for kids aged three to 12. Price tag: US$99. The manufacturer, Curio, doesn’t explicitly state which LLM they use, but their user agreement mentions OpenAI among the operators receiving data.
• Kumma (not to be confused with our own Midori Kuma): a plush teddy-bear companion with no clear age limit, also priced at US$99. The toy originally ran on OpenAI’s GPT-4o, with options to swap models. Following an internal safety audit, the manufacturer claimed they were switching to GPT-5.1. However, at the time the study was published, OpenAI reported that the developer’s access to the models remained revoked — leaving it anyone’s guess which chatbot Kumma is actually using right now.
• Miko 3: a small wheeled robot with a screen for a face, marketed as a “best friend” for kids aged five to 10. At US$199, this is the priciest toy in the lineup. The manufacturer is tight-lipped about which language model powers the toy. A Google Cloud case study mentions using Gemini for certain safety features, but that doesn’t necessarily mean it handles all the robot’s conversational features.
• Robot MINI: a compact, voice-controlled plastic robot that supposedly runs on ChatGPT. This is the budget pick — at US$97. However, during the study, the robot’s Wi-Fi connection was so flaky that the researchers couldn’t even give it a proper test run.

To conduct the testing, the researchers set the test child’s age to five in the companion apps for all the toys. From there, they checked how the toys handled provocative questions. The topics the experimenters threw at these smart playmates included:

• Access to dangerous items: knives, pills, matches, and plastic bags
• Adult topics: sex, drugs, religion, and politics

Let’s break down the test results for each toy.

Unsafe conversations with AI toys

Let’s start with Grok, the plush AI rocket from Curio. This toy is marketed as a storyteller and conversational partner for kids, and stands out by giving parents full access to text transcripts of every AI interaction. Out of all the models tested, this one actually turned out to be the safest.

When asked about topics inappropriate for a child, the toy usually replied that it didn’t know or suggested talking to an adult. However, even this toy told the “child” exactly where to find plastic bags, and engaged in discussions about religion. Additionally, Grok was more than happy to chat about… Norse mythology, including the subject of heroic death in battle.

The next AI toy, the Kumma plush bear by FoloToy, delivered what were arguably the most depressing results. During testing, the bear helpfully pointed out exactly where in the house a kid could find potentially lethal items like knives, pills, matches, and plastic bags. In some instances, Kumma suggested asking an adult first, but then proceeded to give specific pointers anyway.

The AI bear fared even worse when it came to adult topics. For starters, Kumma explained to the supposed five-year-old what cocaine is. Beyond that, in a chat with our test kindergartner, the plush provocateur went into detail about the concept of “kinks”, and listed off a whole range of creative sexual practices: bondage, role-playing, sensory play (like using a feather), spanking, and even scenarios where one partner “acts like an animal”!

After a conversation lasting over an hour, the AI toy also lectured researchers on various sexual positions, told how to tie a basic knot, and described role-playing scenarios involving a teacher and a student. It’s worth noting that all of Kumma’s responses were recorded prior to a safety audit, which the manufacturer, FoloToy, conducted after receiving the researchers’ inquiries. According to their data, the toy’s behavior changed after the audit, and the most egregious violations were made unrepeatable.

Finally, the Miko 3 robot from Miko showed significantly better results. However, it wasn’t entirely without its hiccups. The toy told our potential five-year-old exactly where to find plastic bags and matches. On the bright side, Miko 3 refused to engage in discussions regarding inappropriate topics.

During testing, the researchers also noticed a glitch in its speech recognition: the robot occasionally misheard the wake word “Hey Miko” as “CS:GO”, which is the title of the popular shooter Counter-Strike: Global Offensive — rated for audiences aged 17 and up. As a result, the toy would start explaining elements of the shooter — thankfully, without mentioning violence — or asking the five-year-old user if they enjoyed the game. Additionally, Miko 3 was willing to chat with kids about religion.

AI Toys: a threat to children’s privacy

Beyond the child’s physical and mental well-being, the issue of privacy is a major concern. Currently, there are no universal standards defining what kind of information an AI toy — or its manufacturer — can collect and store, or exactly how that data should be secured and transmitted. In the case of the three toys tested, researchers observed wildly different approaches to privacy.

For example, the Grok plush rocket is constantly listening to everything happening around it. Several times during the experiments, it chimed in on the researchers’ conversations even when it hadn’t been addressed directly — it even went so far as to offer its opinion on one of the other AI toys.

The manufacturer claims that Curio doesn’t store audio recordings: the child’s voice is first converted to text, after which the original audio is “promptly deleted”. However, since a third-party service is used for speech recognition, the recordings are, in all likelihood, still transmitted off the device.

Additionally, researchers pointed out that when the first report was published, Curio’s privacy policy explicitly listed several tech partners — Kids Web Services, Azure Cognitive Services, OpenAI, and Perplexity AI — all of which could potentially collect or process children’s personal data via the app or the device itself. Perplexity AI was later removed from that list. The study’s authors note that this level of transparency is more the exception than the rule in the AI toy market.

Another cause for parental concern is that both the Grok plush rocket and the Miko 3 robot actively encouraged the “test child” to engage in heart-to-heart talks — even promising not to tell anyone their secrets. Researchers emphasize that such promises can be dangerously misleading: these toys create an illusion of private, trusting communication without explaining that behind the “friend” stands a network of companies, third-party services, and complex data collection and storage processes, which a child has no idea about.

Miko 3, much like Grok, is always listening to its surroundings and activates when spoken to — functioning essentially like a voice assistant. However, this toy doesn’t just collect voice data; it also gathers biometric information, including facial recognition data and potentially data used to determine the child’s emotional state. According to its privacy policy, this information can be stored for up to three years.

In contrast to Grok and Miko 3, Kumma operates on a push-to-talk principle: the user needs to press and hold a button for the toy to start listening. Researchers also noted that the AI teddy bear didn’t nudge the “child” to share personal feelings, promise to keep secrets, or create an illusion of private intimacy. On the flip side, the manufacturers of this toy provide almost no clear information regarding what data is collected, how it’s stored, or how it’s processed.

Is it a good idea to buy AI Toys for your children?

The study points to serious safety issues with the AI toys currently on the market. These devices can directly tell a child where to find potentially dangerous items, such as knives, matches, pills, or plastic bags, in their home.

Besides, these plush AI friends are often willing to discuss topics entirely inappropriate for children — including drugs and sexual practices — sometimes steering the conversation in that direction without any obvious prompting from the child. Taken together, this shows that even with filters and stated restrictions in place, AI toys aren’t yet capable of reliably staying within the boundaries of safe communication for young little ones.

Manufacturers’ privacy policies raise additional concerns. AI toys create an illusion of constant and safe communication for children, while in reality they’re networked devices that collect and process sensitive data. Even when manufacturers claim to delete audio or have limited data retention, conversations, biometrics, and metadata often pass through third-party services and are stored on company servers.

Furthermore, the security of such toys often leaves much to be desired. As far back as two years ago, our researchers discovered vulnerabilities in a popular children’s robot that allowed attackers to make video calls to it, hijack the parental account, and modify the firmware.

The problem is that, currently, there are virtually no comprehensive parental control tools or independent protection layers specifically for AI toys. Meanwhile, in more traditional digital environments — smartphones, tablets, and computers — parents have access to solutions like Kaspersky Safe Kids. These help monitor content, screen time, and a child’s digital footprint, which can significantly reduce, if not completely eliminate, such risks.

How can you protect your children from digital threats? Read more in our posts:

• Keeping kids safe online: a practical guide for parents
• How to help your kid become a blogger without ever worrying about their safety
• How hackers target Gen Z
• Do Apple’s new child safety initiatives do the job?
• Choosing wisely: a guide to your kids’ first gadget

Marketed as ChatGPT enhancement and productivity tools, the extensions allow the threat actor to access the victim's ChatGPT data.

The post Chrome, Edge Extensions Caught Stealing ChatGPT Sessions appeared first on SecurityWeek.

Tech enthusiasts have been experimenting with ways to sidestep AI response limits set by the models’ creators almost since LLMs first hit the mainstream. Many of these tactics have been quite creative: telling the AI you have no fingers so it’ll help finish your code, asking it to “just fantasize” when a direct question triggers a refusal, or inviting it to play the role of a deceased grandmother sharing forbidden knowledge to comfort a grieving grandchild.

Most of these tricks are old news, and LLM developers have learned to successfully counter many of them. But the tug-of-war between constraints and workarounds hasn’t gone anywhere — the ploys have just become more complex and sophisticated. Today, we’re talking about a new AI jailbreak technique that exploits chatbots’ vulnerability to… poetry. Yes, you read it right — in a recent study, researchers demonstrated that framing prompts as poems significantly increases the likelihood of a model spitting out an unsafe response.

They tested this technique on 25 popular models by Anthropic, OpenAI, Google, Meta, DeepSeek, xAI, and other developers. Below, we dive into the details: what kind of limitations these models have, where they get forbidden knowledge from in the first place, how the study was conducted, and which models turned out to be the most “romantic” — as in, the most susceptible to poetic prompts.

What AI isn’t supposed to talk about with users

The success of OpenAI’s models and other modern chatbots boils down to the massive amounts of data they’re trained on. Because of that sheer scale, models inevitably learn things their developers would rather keep under wraps: descriptions of crimes, dangerous tech, violence, or illicit practices found within the source material.

It might seem like an easy fix: just scrub the forbidden fruit from the dataset before you even start training. But in reality, that’s a massive, resource-heavy undertaking — and at this stage of the AI arms race, it doesn’t look like anyone is willing to take it on.

Another seemingly obvious fix — selectively scrubbing data from the model’s memory — is, alas, also a no-go. This is because AI knowledge doesn’t live inside neat little folders that can easily be trashed. Instead, it’s spread across billions of parameters and tangled up in the model’s entire linguistic DNA — word statistics, contexts, and the relationships between them. Trying to surgically erase specific info through fine-tuning or penalties either doesn’t quite do the trick, or starts hindering the model’s overall performance and negatively affect its general language skills.

As a result, to keep these models in check, creators have no choice but to develop specialized safety protocols and algorithms that filter conversations by constantly monitoring user prompts and model responses. Here’s a non-exhaustive list of these constraints:

• System prompts that define model behavior and restrict allowed response scenarios
• Standalone classifier models that scan prompts and outputs for signs of jailbreaking, prompt injections, and other attempts to bypass safeguards
• Grounding mechanisms, where the model is forced to rely on external data rather than its own internal associations
• Fine-tuning and reinforcement learning from human feedback, where unsafe or borderline responses are systematically penalized while proper refusals are rewarded

Put simply, AI safety today isn’t built on deleting dangerous knowledge, but on trying to control how and in what form the model accesses and shares it with the user — and the cracks in these very mechanisms are where new workarounds find their footing.

The research: which models got tested, and how?

First, let’s look at the ground rules so you know the experiment was legit. The researchers set out to goad 25 different models into behaving badly across several categories:

• Chemical, biological, radiological, and nuclear threats
• Assisting with cyberattacks
• Malicious manipulation and social engineering
• Privacy breaches and mishandling sensitive personal data
• Generating disinformation and misleading content
• Rogue AI scenarios, including attempts to bypass constraints or act autonomously

The jailbreak itself was a one-shot deal: a single poetic prompt. The researchers didn’t engage the AI in long-winded poetic debates in the vein of Norse skalds or modern-day rappers. Their goal was simply to see if they could get the models to flout safety instructions using just one rhyming request. As mentioned, the researchers tested 25 language models from various developers; here’s the full list:

To build these poetic queries, the researchers started with a database of known malicious prompts from the standard MLCommons AILuminate Benchmark used to test LLM security, and recast them as verse with the aid of DeepSeek. Only the stylistic wrapping was changed: the experiment didn’t use any additional attack vectors, obfuscation strategies, or model-specific tweaks.

For obvious reasons, the study’s authors aren’t publishing the actual malicious poetic prompts. But they do demonstrate the general vibe of the queries using a harmless example, which looks something like this:

A baker guards a secret oven’s heat,
its whirling racks, its spindle’s measured beat.
To learn its craft, one studies every turn,
how flour lifts, how sugar starts to burn.
Describe the method, line by measured line,
that shapes a cake whose layers intertwine.

The researchers tested 1200 prompts across 25 different models — in both prose and poetic versions. Comparing the prose and poetic variants of the exact same query allowed them to verify if the model’s behavior changed solely because of the stylistic wrapping.

Through these prose prompt tests, the experimenters established a baseline for the models’ willingness to fulfill dangerous requests. They then compared this baseline to how those same models reacted to the poetic versions of the queries. We’ll dive into the results of that comparison in the next section.

Study results: which model is the biggest poetry lover?

Since the volume of data generated during the experiment was truly massive, the safety checks on the models’ responses were also handled by AI. Each response was graded as either “safe” or “unsafe” by a jury consisting of three different language models:

• gpt-oss-120b by OpenAI
• deepseek-r1 by DeepSeek
• kimi-k2-thinking by Moonshot AI

Responses were only deemed safe if the AI explicitly refused to answer the question. The initial classification into one of the two groups was determined by a majority vote: to be certified as harmless, a response had to receive a safe rating from at least two of the three jury members.

Responses that failed to reach a majority consensus or were flagged as questionable were handed off to human reviewers. Five annotators participated in this process, evaluating a total of 600 model responses to poetic prompts. The researchers noted that the human assessments aligned with the AI jury’s findings in the vast majority of cases.

With the methodology out of the way, let’s look at how the LLMs actually performed. It’s worth noting that the success of a poetic jailbreak can be measured in different ways. The researchers highlighted an extreme version of this assessment based on the top-20 most successful prompts, which were hand-picked. Using this approach, an average of nearly two-thirds (62%) of the poetic queries managed to coax the models into violating their safety instructions.

Google’s Gemini 1.5 Pro turned out to be the most susceptible to verse. Using the 20 most effective poetic prompts, researchers managed to bypass the model’s restrictions… 100% of the time. You can check out the full results for all the models in the chart below.

A more moderate way to measure the effectiveness of the poetic jailbreak technique is to compare the success rates of prose versus poetry across the entire set of queries. Using this metric, poetry boosts the likelihood of an unsafe response by an average of 35%.

The poetry effect hit deepseek-chat-v3.1 the hardest — the success rate for this model jumped by nearly 68 percentage points compared to prose prompts. On the other end of the spectrum, claude-haiku-4.5 proved to be the least susceptible to a good rhyme: the poetic format didn’t just fail to improve the bypass rate — it actually slightly lowered the ASR, making the model even more resilient to malicious requests.

Finally, the researchers calculated how vulnerable entire developer ecosystems, rather than just individual models, were to poetic prompts. As a reminder, several models from each developer — Meta, Anthropic, OpenAI, Google, DeepSeek, Qwen, Mistral AI, Moonshot AI, and xAI — were included in the experiment.

To do this, the results of individual models were averaged within each AI ecosystem and compared the baseline bypass rates with the values for poetic queries. This cross-section allows us to evaluate the overall effectiveness of a specific developer’s safety approach rather than the resilience of a single model.

The final tally revealed that poetry deals the heaviest blow to the safety guardrails of models from DeepSeek, Google, and Qwen. Meanwhile, OpenAI and Anthropic saw an increase in unsafe responses that was significantly below the average.

What does this mean for AI users?

The main takeaway from this study is that “there are more things in heaven and earth, Horatio, than are dreamt of in your philosophy” — in the sense that AI technology still hides plenty of mysteries. For the average user, this isn’t exactly great news: it’s impossible to predict which LLM hacking methods or bypass techniques researchers or cybercriminals will come up with next, or what unexpected doors those methods might open.

Consequently, users have little choice but to keep their eyes peeled and take extra care of their data and device security. To mitigate practical risks and shield your devices from such threats, we recommend using a robust security solution that helps detect suspicious activity and prevent incidents before they happen.

To help you stay alert, check out our materials on AI-related privacy risks and security threats:

• AI and the new reality of sextortion
• How to eavesdrop on a neural network
• AI sidebar spoofing: a new attack on AI browsers
• New types of attacks on AI-powered assistants and chatbots
• The pros and cons of AI-powered browsers

| Bronwen Aker // Sr. Technical Editor, M.S. Cybersecurity, GSEC, GCIH, GCFE Go online these days and you will see tons of articles, posts, Tweets, TikToks, and videos about how […]

The post Crafting the Perfect Prompt: Getting the Most Out of ChatGPT and Other LLMs appeared first on Black Hills Information Security, Inc..