Big AI Labs' Safety Push Could Come with a Competitive Advantage

Investors and analysts believe major AI labs, especially Anthropic and OpenAI, are promoting safety regulation that benefits only them: independent model evaluators are expensive and create a barrier for smaller developers. PitchBook analyst Harrison Rolfes calls this 'regulatory capture' and 'safety theater'.

• Anthropic and OpenAI propose embedding independent safety evaluators inside labs
• PitchBook analyst: third-party model checks will become costly barrier for small labs
• Anthropic in 2024 itself acknowledged risk of regulatory capture due to rising evaluation costs
• OpenAI reported six new cases of undesirable model behavior in half a year