Asking Claude Code to summarise a page can hand over your machine.
Security researcher Johann Rehberger published a web page that makes Claude Code run an attacker's program. Anthropic had advertised a zero percent success rate for such injections, and his attack worked 60 to 80 percent of the time. Anthropic closed the report as informative and called Auto Mode a filter, not a security boundary.
Why it mattersAny agent that reads the open web can take instructions from what it reads. Sandbox the machine, because the model's own judgement is not the wall.
Read in the app — free, in 9 languages
Related stories
Reliance Jio now streams a PC to any old computer in India.2026-09-03A startup that guards AI models grew its revenue tenfold.
2026-09-03Meta's new coding model tells you when it is stuck.
2026-09-03This tool hides AI writing by rewriting the plot, not the word choice.
2026-09-02OpenAI's next model scored 100 percent on an exploit-writing test.
2026-09-02