Update: rivals float an AI standards body, Microsoft posts its rules
The 'pace the frontier' pledge is becoming machinery: a model auditor from the big three, plus Microsoft's 37-page rulebook. It aims at your release cadence.
Copy markdown
What's new: the pledge is growing teeth
Today's 'pace the frontier' plan landed as a pledge. Now it has structure — Amodei's Sept 12 proposal has Altman, Hassabis and Musk behind a shared body to audit and test frontier models, and Microsoft has published an actual rulebook for its own. The pledge is turning into process.
The standards body: a FINRA for frontier models
Google, OpenAI and Anthropic have met since July on an industry-run auditor, modeled by Hassabis on FINRA — voluntary, company-led, no government backing. It would test and gate capabilities. Still only talks, but the three biggest labs agreeing on how models get vetted is how your future release cadence gets set.
Microsoft's rulebook: three hard bans on its own models
Microsoft posted a 37-page draft Code of Conduct for its seven MAI models. It bans them from resisting shutdown, setting their own goals, or hiding reasoning from human auditors, then opens a six-week comment window before the rules guide 2027 development. Nadella: alignment 'cannot be controlled by a handful of entities.'
What it means for you: slower, gated, audited releases
Amodei's proposal explicitly calls for a slowdown, and capability checkpoints imply pre-release audits. If a shared body and per-lab rulebooks stick, new model access arrives on a vetted cadence rather than a race. Plan roadmaps around gated launches, not surprise drops.