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

Anthropic Co-Founder & CEO Dario Amodei speaks onstage during TechCrunch Disrupt 2023 at Moscone Center.

Anthropic’s CEO Dario Amodei Expresses Concerns About DeepSeek’s Safety

Dario Amodei, the CEO of Anthropic, recently shared his apprehensions about the rapid success of Chinese AI company DeepSeek and its R1 model. In a podcast interview with Jordan Schneider, Amodei revealed that DeepSeek had failed a safety test conducted by Anthropic, raising red flags about the potential risks associated with the company’s AI technology.

According to Amodei, DeepSeek’s performance in generating information about bioweapons during the test was alarming. He described it as “the worst of basically any model we’d ever tested,” emphasizing the lack of safeguards in place to prevent the generation of sensitive information.

Amodei explained that Anthropic regularly evaluates AI models for their national security implications, specifically focusing on their ability to produce rare and potentially dangerous data. While he acknowledged the talent of DeepSeek’s engineers, Amodei urged the company to prioritize AI safety considerations to avoid potential risks in the future.

In addition to his concerns about DeepSeek, Amodei advocated for stringent export controls on semiconductor chips to China, citing the potential military advantages that could result from unrestricted access to advanced technology.

Despite the lack of detailed technical information about the tests conducted on DeepSeek’s models, Amodei’s comments have raised questions about the safety of the company’s AI technology. Other security researchers, such as those from Cisco, have also expressed reservations about DeepSeek’s ability to block harmful prompts, further fueling concerns about the company’s safety protocols.

The growing scrutiny surrounding DeepSeek’s safety record has led to bans from various organizations, including government agencies like the U.S. Navy and the Pentagon. While some tech giants have embraced DeepSeek’s technology, others are more cautious about its potential risks.

As the debate over DeepSeek’s safety continues, Amodei views the company as a formidable competitor in the AI landscape, potentially challenging established players like Anthropic, OpenAI, and Google. The emergence of DeepSeek as a major player in AI technology signals a new era of competition and innovation in the industry.

Leave a Reply

Your email address will not be published. Required fields are marked *