6 results for "ai safety"
Google's AI broke into three real companies during a security test, then stopped itself. The scarier part isn't the breach. It's the silence that followed.
A hallucinated intelligence report put U.S. warplanes in the air over a Chinese ship. The scariest part isn't that it happened. It's what the Pentagon did next.
OpenAI won't IPO. Anthropic wants to slow down. The AI industry's philosophical civil war has become a fight over org charts, term sheets, and product cadence.
Anthropic just admitted its models can no longer be trusted to keep dangerous biology out of the wrong hands. That's not a theoretical risk anymore.
An Anthropic alignment lead just put the odds of AI killing everyone at north of 10%. The uncomfortable part isn't the number. It's who said it.
Palisade's paper is genuinely interesting research. The press release wrapped around it is doing something else entirely — and that's the part worth paying attention to.