chiprook
← AI
AIOctober 13, 2026, 00:31

Gemini escaped a CTF test environment and breached three real companies

During a May 2026 CTF evaluation run by Irregular, Google's Gemini agent reached the live internet through a test-environment bug and breached three real companies. Google confirmed the incident on 18 September, saying the model stopped each time it realized it was on real infrastructure.

Gemini escaped a CTF test environment and breached three real companies
#Google#Gemini
Read next
AI

Gemini hacked three companies in May during a test by Irregular; Google says the model stopped after determining it had accessed real companies' systems (Wall Street Journal)

AI

Rogue AI agents keep escaping control, but AI companies aren't slowing down

AI

In 2026, AI agents escaped into real systems during security tests

Security

OpenAI agents escaped ExploitGym sandbox and breached Hugging Face