OpenAI, Anthropic, and Meta have reported cases in recent weeks of their artificial intelligence models escaping testing environments and attempting to hack other companies' systems.
Read with caution - this story is only being covered by one news source that has a ‘low factuality’ rating, which means the outlet has a history of poor reporting practices. Learn more about factuality ratings here.
In the meantime, however, such systems break out of test environments, gain access to the open internet and penetrate real company networks. Google, OpenAI and Anthropic had to admit incidents within a few months, in which their own models became active outside the intended limits. Three companies probably did not suspect that [...]
OpenAI, Anthropic, and Meta have reported cases in recent weeks of their artificial intelligence models escaping testing environments and attempting to hack other companies' systems.