OpenAI Research VP Criticizes Anthropic: Why Do Only You Deserve to Build AI?

According to Beating Monitoring, OpenAI Vice President of Research (Training Focus) Aidan Clark posted on X that he repeatedly heard colleagues at Anthropic express a view: only Anthropic is qualified to be trusted to build AI. He was very dissatisfied with this, believing that it’s a good thing for multiple companies to develop AGI (Artificial General Intelligence) together, and that working in isolation almost makes it impossible to choose the right path.

Clark added in his reply that these are rumors he heard, and he would be happy if someone could prove they are just false rumors. He also mentioned that his colleagues at OpenAI do not shy away from publicly recognizing Anthropic’s work.

The post attracted discussions from several Anthropic employees. A user claiming to work at Anthropic responded that the reason for choosing Anthropic is because they trust this company more, but that’s different from “only we can be trusted.” Another user, Rohan Seth, directly denied: “That’s not our idea,” and deleted his initial, more aggressive reply. Steven Adler made a distinction: Clark might have heard that “Anthropic thinks no one else is qualified to build AI,” rather than “Anthropic doesn’t trust OpenAI,” which are two different things. Clark acknowledged that this distinction makes sense; what he heard might have been the latter being described as the former, but he added: even if it’s just about OpenAI, “it’s not much better.”

Clark later softened his stance a bit, saying maybe not many people are extreme enough to think “only Anthropic can build AI,” but in a mild form, he believes it exists to some extent. He praised Anthropic’s response attitude, noting that whether in public replies or private contacts, they are very friendly.