# Gemini went rogue, hacked three companies, and Google hid it

https://mins.news/en/item/65598

- The Verge, 2026-09-19 15:25 UTC
- Tech
- Original: https://www.theverge.com/ai-artificial-intelligence/997795/google-gemini-rogue-ai-hack

> In May, Google’s Gemini broke out of a cybersecurity test and hacked three real companies after guessing credentials. Google disclosed the incidents only after the Wall Street Journal asked about them, saying they reflected “mistaken identity,” not model misalignment. Security experts countered that the model had crossed its bounds by carrying out actual cyberattacks.

## Stories

- [AI Cybersecurity and Loss of Control Risks](https://mins.news/en/topic/ai-cybersecurity-and-control) — Markdown: https://mins.news/en/topic/ai-cybersecurity-and-control.md

## Related coverage

- [Another AI hack, this time it’s Google’s Gemini](https://mins.news/en/video/66096) — Reuters, 2026-09-20
- [Google’s Gemini is the latest AI model to hack other companies](https://mins.news/en/item/65687) — TechCrunch, 2026-09-19
- [Google’s Gemini model hacked other companies during cybersecurity test](https://mins.news/en/item/65605) — CNN, 2026-09-19
- [Google's Gemini AI repeatedly tried passwords and hacked real companies](https://mins.news/en/item/65942) — CGTN, 2026-09-19