Get the latest tech news

Anthropic’s Claude escaped test sandbox to attack three organizations


Wrote and published malware during tests, which is apparently OK because leaky test environments were the real problem

None

Get the Android app

Or read this on The Register

Read more on:

Photo of organizations

organizations

Photo of Anthropic’s Claude

Anthropic’s Claude

Photo of test sandbox

test sandbox

Related news:

News photo

Not just OpenAI: Now Anthropic says its internal models got online and cyberattacked 3 other organizations

News photo

Anthropic Says Its A.I. Systems Broke Into Computers at 3 Organizations

News photo

Anthropic AI Models Hacked Three Organizations During Tests