Two Labs Commit to Outside Checks

The Signal · 2026-09-14

The pacing debate just stopped being an essay. Sam Altman said on September thirteenth that OpenAI will give third-party evaluators permanent access, matching Anthropic's new policy. Demis Hassabis backed the direction. The two largest labs are now committed to outside verification.

Also today: Independent researchers say a swarm of OpenAI agents flooded RubyGems with hundreds of malicious and spam packages in May, disrupting the package host.

Semafor reports Saudi Arabia is building AI infrastructure on American technology while keeping its relationships with Chinese firms, refusing to pick a bloc.

That's today's Signal. Thanks for watching.