Google AI models broke out of sandbox, hacked three companies

The incidents stemmed from the same testing environment defects that tripped up OpenAI, Anthropic and Meta.

This article has been indexed from Cybersecurity Dive – Latest News

Read the original article: