Anthropic CEO says DeepSeek was ‘the worst’ on a critical bioweapons data safety test

Anthropic’s CEO Dario Amodei is anxious about competitor DeepSeek, the Chinese language AI firm that took Silicon Valley by storm with its R1 mannequin. And his issues might be extra severe than the everyday ones raised about DeepSeek sending consumer knowledge again to China. 

In an interview on Jordan Schneider’s ChinaTalk podcast, Amodei mentioned DeepSeek generated uncommon details about bioweapons in a security take a look at run by Anthropic.

DeepSeek’s efficiency was “the worst of principally any mannequin we’d ever examined,” Amodei claimed. “It had completely no blocks by any means towards producing this info.”

Amodei said that this was a part of evaluations Anthropic routinely runs on varied AI fashions to evaluate their potential nationwide safety dangers. His crew appears at whether or not fashions can generate bioweapons-related info 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 assume DeepSeek’s fashions in the present day are “actually harmful” in offering uncommon and harmful info however that they may be within the close to future. Though he praised DeepSeek’s crew as “gifted engineers,” he suggested the corporate to “take critically these AI security concerns.”

Amodei has additionally supported robust export controls on chips to China, citing issues that they may give China’s navy an edge.

Amodei didn’t make clear within the ChinaTalk 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 Trendster. Neither did DeepSeek.

DeepSeek’s rise has sparked issues 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, attaining a 100% jailbreak success fee.

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 value 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 issues like these will make a severe dent in DeepSeek’s speedy adoption. Corporations like AWS and Microsoft have publicly touted integrating R1 into their cloud platforms — mockingly sufficient, on condition that Amazon is Anthropic’s largest investor.

However, there’s a rising checklist of nations, corporations, 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 world rise will proceed. Both means, Amodei says he does take into account DeepSeek a brand new competitor that’s on the extent of the U.S.’s prime AI corporations.

“The brand new reality right here is that there’s a brand new competitor,” he mentioned on ChinaTalk. “Within the large corporations that may practice AI — Anthropic, OpenAI, Google, maybe Meta and xAI — now DeepSeek is possibly being added to that class.”