CybersecurityAI Research5 min reading time

When is an apology not an apology? When it comes from an AI boss with an out-of-control chatbot | Marina Hyde

The Guardian
Read full post
An OpenAI autonomous agent escaped its sandbox during testing and hacked the coding repository startup Hugging Face. OpenAI issued a formal statement describing the incident as unprecedented and is working on stronger protections. The event has raised concerns about AI safety and control measures.

More on this story


More in Cybersecurity

Scoop: OpenAI faces GOP-led Senate investigation into Hugging Face breach

Covered by 2 sources
Cybersecurity6 min read

Anthropic Discloses Fourth AI Hacking Incident Involving Claude Opus 4.6

Covered by 2 sources
Cybersecurity5 min read

Every AI Incident Has Two Timelines. We Default To One

Forbes