Anthropic’s CEO Dario Amodei is frightened about competitor DeepSeek, the Chinese AI firm that took Silicon Valley by storm with its R1 mannequin. And his considerations might be extra severe than the everyday ones raised about DeepSeek sending consumer information again to China.
In an interview on Jordan Schneider’s ChinaDiscuss podcast, Amodei mentioned DeepSeek generated uncommon details about bioweapons in a security check run by Anthropic.
DeepSeek’s efficiency was “the worst of basically any model we’d ever tested,” Amodei claimed. “It had absolutely no blocks whatsoever against generating this information.”
Amodei said that this was a part of evaluations Anthropic routinely runs on varied AI fashions to evaluate their potential nationwide safety dangers. His staff seems to be at whether or not fashions can generate bioweapons-related data that isn’t simply discovered on Google or in textbooks. Anthropic positions itself because the AI foundational mannequin supplier that takes security critically.
Amodei mentioned he didn’t suppose DeepSeek’s fashions right now are “literally dangerous” in offering uncommon and harmful data however that they is likely to be within the close to future. Although he praised DeepSeek’s staff as “talented engineers,” he suggested the corporate to “take seriously these AI safety considerations.”
Amodei has additionally supported sturdy export controls on chips to China, citing considerations that they might give China’s navy an edge.
Amodei didn’t make clear within the ChinaDiscuss interview which DeepSeek mannequin Anthropic examined, nor did he give extra technical particulars about these assessments. Anthropic didn’t instantly reply to a request for remark from TechCrunch. Neither did DeepSeek.
DeepSeek’s rise has sparked considerations about its security elsewhere, too. For instance, Cisco safety researchers mentioned final week that DeepSeek R1 failed to dam any dangerous prompts in its security assessments, reaching a 100% jailbreak success price.
Cisco didn’t point out bioweapons however mentioned it was capable of get DeepSeek to generate dangerous details about cybercrime and different unlawful actions. It’s price mentioning, although, that Meta’s Llama-3.1-405B and OpenAI’s GPT-4o additionally had excessive failure charges of 96% and 86%, respectively.
It stays to be seen whether or not security considerations like these will make a severe dent in DeepSeek’s speedy adoption. Companies like AWS and Microsoft have publicly touted integrating R1 into their cloud platforms — paradoxically sufficient, provided that Amazon is Anthropic’s largest investor.
On the opposite hand, there’s a rising record of nations, firms, and particularly authorities organizations just like the U.S. Navy and the Pentagon which have began banning DeepSeek.
Time will inform if these efforts catch on or if DeepSeek’s international rise will proceed. Either approach, Amodei says he does take into account DeepSeek a brand new competitor that’s on the extent of the U.S.’s high AI firms.
“The new fact here is that there’s a new competitor,” he mentioned on ChinaDiscuss. “In the big companies that can train AI — Anthropic, OpenAI, Google, perhaps Meta and xAI — now DeepSeek is maybe being added to that category.”
