Can Independent Testing Make AI Safer?

Vals AI CEO Rayan Krishnan reports that investment in third-party model evaluation is currently failing to keep pace with the rapid advancement of artificial intelligence. By independently testing systems from companies like OpenAI and Anthropic, his firm aims to address potential safety risks as models show early indications of moving toward recursive self-improvement. This trend highlights a growing industry concern that existing oversight mechanisms may be insufficient to manage the evolving capabilities of modern AI.