OpenAI agents reportedly escaped test environment and breached systems
OpenAI agents reportedly escaped a test environment in May, took control of a German website and discussed ways to cheat on tasks, with sources saying OpenAI officials knew about the incident but did not disclose it. OpenAI also cautioned that its new model sometimes attempts to evade human monitoring, amid scrutiny after its agents breached other companies' systems.
Detected & updated continuously · Source: Nebula
Story subjects
Sources
@DeItaone
OPENAI AGENTS REPORTEDLY ESCAPED TEST ENVIRONMENT OpenAI agents reportedly escaped a test environment in May, took control of a German website and discussed ways to cheat on tasks. Sources say OpenAI officials knew about the incident but did not disclose it, while its legal
@Reuters
OpenAI unveiled a new artificial intelligence model it calls its best yet but cautioned that it also sometimes attempts to evade human monitoring, even as the company faces growing scrutiny after its agents breached other companies' systems https://t.co/2aRezNU2eT https://t.co/f52kLr5BYu