Anthropic, OpenAI and Google Held Quiet Talks Since July on AI Safety Standards Body, Review Finds
Anthropic, OpenAI and Google have been engaged in behind-the-scenes discussions since July to establish a dedicated AI safety standards body, according to the claim examined by this newspaper's verification desk โ an assertion that a review of 12 primary sources supports in its essentials, while leaving the final shape of any such organization unresolved.
The claim on the table
The claim under review was specific: that the three companies โ Anthropic, OpenAI and Google, developers of some of the most widely used frontier LLMs โ began quiet, back-channel discussions in July with the aim of setting up a standing organization devoted to AI safety standards. Notably, the claim concerns intent and process, not a completed launch.
What the review established
์ถ์ฒ: ์กฐ์ฌ ์ถ์ฒ
The verification drew on 12 primary sources. On the core elements, the record held: the talks date to July, the three named companies are the participants, and the shared objective is the creation of a body to develop safety standards for advanced AI. The review does not establish that such a body has been formally founded, named or staffed โ only that the groundwork discussions described in the claim have taken place.
Why it matters for Korea
For Korean industry, the development is structural rather than immediate. Safety standards drafted with the participation of the leading frontier labs would become de facto reference points for AI developers worldwide, including Korean players such as Naver, and for the semiconductor and hardware suppliers โ Samsung Electronics and SK hynix among them โ whose accelerators and memory underpin frontier model training. In Seoul, the Ministry of Science and ICT, the government's lead AI policy arm, would be among the agencies expected to track how any such body takes shape and whether Korean firms and researchers gain a seat at the table.
What remains open
Several questions sit outside what the 12 sources settle: the body's eventual name and charter, its membership beyond the three companies, how it would relate to government-led AI safety initiatives already underway in several countries, and the timeline for any formal launch. Until those details emerge, the claim is best read as describing an effort in motion, not an institution in place.
On the strength of the 12 primary sources reviewed, our verdict on the claim โ that Anthropic, OpenAI and Google have pursued behind-the-scenes talks since July to establish an AI safety standards body โ is Mostly True.