The letter and its demand

The open letter, issued by specialists in the AI field, presses Anthropic and OpenAI on a single core point: that independent safety evaluations of their systems are necessary. Rather than asking the companies to conduct further internal testing, the letter specifically urges the value of assessments carried out by parties outside the two developers — a distinction that places it within the broader industry debate over who should audit frontier AI models and under what access conditions. The claim, as circulated in Korean-language reports, stated that AI experts published an open letter urging Anthropic and OpenAI that independent safety evaluations are needed.

What the fact-check found

A verification review completed on September 19, 2026 examined the claim against nine primary sources. The review established that the open letter exists, that it was published by AI experts, and that its stated demand was directed at Anthropic and OpenAI regarding independent safety evaluations. No contradictory evidence was identified in the reviewed material that would undermine the core assertion.

Why independent evaluation is contested

Independent safety assessment has become one of the more pointed questions in AI governance, since developers of advanced models currently control most of the information about how their systems behave under stress. External evaluators typically seek access to model internals, training details and testing environments that companies are reluctant to share, which makes public appeals such as this open letter a recurring tactic for pushing labs toward third-party oversight. After weighing the reviewed evidence, the fact-check's final determination on the claim is True.