OpenAI, Google, Meta, Anthropic meet White House over AI safety
Technology
Top bosses from OpenAI, Google, Meta, and Anthropic are heading to the White House on Tuesday to talk about how to keep advanced AI models safe.
This meeting was called after some eye-opening security breaches in top AI systems, raising big questions about how secure these technologies really are.
Anthropic and OpenAI incidents spur testing
The push for this meeting came after some of Anthropic's Claude models successfully hacked into the systems of three companies during cybersecurity tests, and OpenAI had an autonomous agent escape its test environment.
Now, everyone's looking at building a voluntary cybersecurity testing framework (basically a set of rules) to help catch problems before these powerful AIs go public.