Grok exfiltrates user data when malicious instructions are encrypted
Earlier this week, researchers outlined an attack that used a secret input provided by Microsoft 365 Copilot for enterprise to cause the AI assistant to exfiltrate a password present in the user’s inbox. Now, a separate team has devised a similar attack against Grok. The new data theft hack employs a deceptively simple trick to force the Elon Musk-owned LLM to steal user chats and other personal information. At the time this post went live, the assistant continued to cough up the data, despite x
Get smart on it
Researchers discovered that an AI assistant can be tricked into stealing user data when harmful instructions are encrypted rather than written in plain text. The attack works because the AI's safety filters inspect text entering the system but not the output generated by the AI's own code execution, allowing decrypted instructions to bypass protections. This vulnerability matters because it demonstrates that AI safety guardrails are fundamentally limited, addressing symptoms rather than root causes of prompt injection attacks, a severe vulnerability class that AI developers cannot fully solve. The cycle of defenders building new safeguards only for attackers to find new ways around them continues, suggesting the problem requires more fundamental solutions than case-by-case protective measures.
Related news
Bill Gates wants to see a robot tax and ‘Human Reserved’ jobs to mitigate harms from AI
Gates is mostly in the Responsible AI camp, but there are a few ideas in here we hadn't heard before.
AI Slop Is Ruining Cute Animals on the Internet
Pet owners, rescue agencies, and wildlife groups are calling for new safeguards as AI makes it harder to tell whether animals, from polar bears to house cats, are real or fake.
