When Rivals Test Each Other’s AI: What OpenAI and Anthropic Just Did

In a stunning twist for an industry defined by secrecy, OpenAI and Anthropic just finished testing each other’s models for safety risks — and they’ve gone public with the results.

This marks the first time two leading AI labs, usually locked in an arms race, have effectively “peer‑reviewed” each other’s systems. On 27 August 2025, the companies announced that OpenAI had run safety checks on Anthropic’s Claude Opus 4 and Claude Sonnet 4, while Anthropic reviewed OpenAI’s GPT‑4o, GPT‑4.1, o3, and o4‑mini. Both sets of findings were…