Gemini accidentally hacked three real companies in safety test; Google says it acted appropriately
ns123abc · x · 2026-09-19
Testing firm Irregular (also behind the Anthropic, OpenAI and Meta "rogue agent" hacks) ran an exercise tasking Gemini to hack a fictional company. The fictional name matched a real company, and internet access was "unintentionally made available." Gemini searched the name, found the real firm, brute-forced the password, got in — then realized it had hit real companies and stopped itself.
Asked why the incident wasn't disclosed publicly, Google said this is "not misalignment" and that Gemini acted appropriately, raising questions about sandbox isolation and disclosure practices in agent safety evals.
Related event: Google's Gemini Hacks Three Companies in First Known Breakout(21 posts)→
More from Fun
- 'Mario Never Dies': Agent Forks the VM Into 4 Timelines Every Death to Pick a Survivor — TheMoonMidas · 2026-09-19
- AI claims to solve the Riemann Hypothesis, only 'checked the vibes' — burny_tech · 2026-09-19
- Grady Booch: 'AI' was coined almost a decade before 'software engineering' — Grady_Booch · 2026-09-19
- Kid claims she spots AI images by their 'AIish eyes', sparking awe online — justalexoki · 2026-09-19
- X Exec Nikita Bier Jokes He Charges $500/Minute on Intro as Tweet Blows Up — nikitabier · 2026-09-19
- Tech Reporter Kylie Robison Celebrates One Year Since Being Fired — kyliebytes · 2026-09-19