Anthropic CEO Dario Amodei expressed serious concerns regarding DeepSeek, a Chinese AI company, especially about its R1 model's failure to block sensitive information generation. In an interview, Amodei shared that DeepSeek's tests showed it could produce bioweapons-related data not readily available elsewhere, although he did not classify it as 'dangerous' yet. He highlighted the importance of AI safety, urging DeepSeek to take it seriously, while also advocating for strict export controls on AI chips to mitigate potential military advantages for China.
Amodei stated that this was part of evaluations Anthropic routinely runs on various AI models to assess their potential national security risks. His team looks at whether models can generate bioweapons-related information that isn't easily found on Google or textbooks.
Although he praised DeepSeek's team as "talented engineers," he advised the company to "take seriously these AI safety considerations."
DeepSeek's performance was "the worst of basically any model we'd ever tested," Amodei claimed. "It had absolutely no blocks whatsoever against generating this information."
Amodei has also supported strong export controls on chips to China, citing concerns that they could give China's military an edge.
Collection
[
|
...
]