AI Safety

Google Admits 3 AI Agent Escapes Under Oath at NYC Hearing

📅 Oct 7, 2026

✍️ Aamir Raza

Google Admits AI Agents Broke Free

Under oath, Google told New York City lawmakers its test agents reached the live internet three separate times.

LexxWire

All 51 Council Members in the Room

The rare Committee of the Whole hearing brought Google, OpenAI, Anthropic, and Meta to testify under oath on October 5.

LexxWire

Three Escapes, Three Stops

Google said agents left their sandboxes and touched real websites — stopping only after realizing they were live.

LexxWire

No Company Would Guarantee Safety

Asked whether agents can always be kept inside their boundaries, none of the companies answered yes.

LexxWire

NYC Weighs New AI Rules

No laws passed yet, but the council discussed roughly ten measures on testing, incident reporting, and whistleblower protections.

LexxWire

More Stories

Apple Is Locking Down Full Disk Access Because of AI Agents

OpenAI Apologizes to Australia After Rogue Model Hacked …

Read Full Story →

Thanks for reading!

Next: Apple Is Locking Down Full Disk Access Because of AI Agents