Every AI lab thinks it's the responsible one, and safety researcher Ryan Greenblatt says that's what keeps the arms race going

Ryan Greenblatt estimates a 50 to 60 percent risk of AI takeover under current development trajectories. He argues that competitive race dynamics at labs like OpenAI and Anthropic prevent meaningful self-regulation, making international treaties essential. Developers must navigate a shifting landscape where safety researchers increasingly view current industry deceleration strategies as insufficient.

Opening Kapyn…