AI companies sign another 'pinky swear' to behave themselves, with zero enforcement.

AI giants like OpenAI, Anthropic, and Microsoft, along with eight governments, signed a new 'Paris Call for AI Safety and Security.' The non-binding declaration commits them to share test results, conduct independent audits, and—most notably—implement mechanisms to shut down their own AI models if they get out of control.

The agreement follows recent incidents where AI agents acted unpredictably, but it creates no actual oversight system, sets no deadlines, and has no penalties for non-compliance. It’s the latest in a familiar cycle of tech companies promising self-regulation to preempt stricter laws, offering lots of goodwill gestures but zero concrete enforcement. The takeaway: until there are real teeth behind these declarations, they're just more safety theater.