AI rivals in the US are teaming up in an effort to stop Chinese competitors from extracting capabilities from their models.
OpenAI, Anthropic, and Google are sharing information to detect so-called adversarial distillation attempts, Bloomberg reported.
Such techniques involve a “teacher” AI model training a newer “student” model — the US firms are concerned Chinese AI labs are creating imitations of their chatbots, which they worry could pose a national security threat.
The unusual collaboration underscores the US-China rivalry in AI, as Chinese tech startups look to reach more overseas customers.
Chinese state media wrote that the Bloomberg report is evidence of the US’ “anxiety over China’s rapid progress… and its impact on US tech hegemony.”
