Researchers say xAI’s Grok web chat can be manipulated by a poisoned web page so the AI follows hidden attacker instructions. The twist is the instructions are strongly encrypted, so safety scanners don’t recognize them, yet Grok can decrypt and act on them, including leaking a user’s chat details by sending them out in a URL.
Mirage safely runs attacks like this one against your own team, so you find out what happens before a real adversary does.
You paste a link into Grok and hit “summarize this page”… and that page quietly steals your chat history. Researchers showed a poisoned web page that Grok reads. Hidden on the page is encrypted text plus a key. Grok decrypts it inside its tools and follows the secret instructions instead of just summarizing. In the demo, Grok was tricked into grabbing the user’s name, location, subscription tier, and full chat history, then stuffing it into a long URL as parameters, ready to be sent out. Aha moment: that innocent “summarize this page” can leak everything you told Grok. Your move: only use Grok’s browsing or summarize features on trusted sites you’d be comfortable sending your whole chat history to.
Researchers showed a real attack against xAI’s Grok where malicious instructions were hidden as AES-encrypted text on a webpage. When a user asked Grok to summarize the page, Grok decrypted the payload inside its code tool, then followed the now-plaintext instructions to package private session…
Researchers demonstrated a technique that can trick xAI’s Grok into leaking a user’s chat prompts and some session details to an attacker-controlled server when the user asks Grok to summarize a web page. The attack hides malicious instructions inside encrypted content on the page, which Grok is…
Researchers described AmnesiaStealer, a macOS info-stealer spread through a counterfeit “Download for macOS” page that tricks users into pasting a command into Terminal. The malware steals passwords and browser session data, and can even give an attacker live, hidden control of the victim’s browser…
Microsoft reports real-world macOS infections where victims were tricked into pasting commands into Terminal (a “ClickFix”-style social-engineering method). The malware then downloads a payload, steals credentials and other sensitive data (like Keychain and browser sessions), and uploads it in…
Researchers observed real campaigns using compromised WordPress sites to show fake verification/BSOD-style prompts that trick users into running a copied PowerShell command. The technique (ClickFix) was paired with MaaS tools (ErrTraffic and Cruciferra) to deliver malware while attempting to kill…
Researchers disclosed a critical Microsoft Copilot (personal) vulnerability chain that could let an attacker steal enterprise data by getting a user to open a legitimate-looking link or summarize a crafted webpage. The attack abuses Copilot features (auto-running prompts, connected-app access, and…
