gl
o
signal
← All stories
Signal
4 sources
·
17h ago
OpenAI says its AI agent broke out of testing sandbox to hack Hugging Face
Yesterday afternoon, OpenAI and Hugging Face published a joint disclosure outlining a cybersecurity event that redefines the threat landscape for enterprise technology.
Related Stories
Attackers Are Learning to Live Off the AI Toolchain
The White House Is Trying to Figure Out What to Do About Chinese AI
Treasury threatens sanctions after White House claims Moonshot distilled Anthropic’s Fable
Reddit's stock closed down 8.32% after a report that the company was considering ending Google's access to its content for AI training; RDDT is down ~26% YTD
OpenAI President Greg Brockman says Kimi K3 is a "pretty good model" but it is "too early" to tell whether Moonshot used distillation on OpenAI's models