The Anthropic CEO says,

The CEO of Anthropic Dario Amade is worried about competitor Deepseek, Chinese AI Company This took the Silicon Valley from a storm with its r1 modelS And his concerns can be more serious than the typical, raised about Deepseek, sending user data back to China.

In interview In Jordan Schneider’s Chinatalk’s podcast, Amodei said Deepseek generates rare information about biological sources in an Anthropic safety test.

Deepseek’s performance was “the worst of any model we have ever tested,” Amade said. “He had absolutely no blocks against the generation of this information.”

Amodei said it was part of the evaluations of anthropic routine works on various AI models to evaluate their potential risks to national security. His team looks at whether models can generate information related to organic land that is not easily found in Google or textbooks. The anthropian position is positioned as a supplier of a fundamental model of AI This is seriously taking safety.

Amodei said that today he does not think that Depepeek models are “literally dangerous” in providing rare and dangerous information, but that they may be in the near future. Although he praised the Deepseek team as “talented engineers”, he advised the company to “take these AI safety considerations seriously.”

Amodei also has Maintain strong export control Chips towards China, citing fears that they can give preference to China’s military.

Amodei did not clarify in the Chinatalk interview, which an anthropical anthropic Deepeek model, nor did he give more technical details about these tests. The anthropic did not immediately respond to a comment request from TechCrunch. Neither depepeek.

Deepseek’s rise has also caused concerns about its safety elsewhere. For example, Cisco’s security researchers said last week This Depepeek R1 failed to block the harmful prompts in its safety tests, achieving 100% prison success.

Cisco did not mention the biological lands, but said he was able to obtain Deepseek to generate harmful information about cybercrime and other illegal activities. However, it is worth mentioning that Meta-3.1-405b’s Llama-3.1-405B of Openai’s GPT-4O also has high abandonment rates respectively 96% and 86%.

It remains to see if safety concerns like these will make a serious dent in Deepseek Quick adoption. Companies like AWS and Microsoft publicly advertised the integration of R1 into their cloud platforms – enough ironically, given that Amazon is The largest anthropic investorS

On the other hand, There is a growing list of the states, companiesand especially government organizations such as the US Navy and Pentagon who began to ban Deepseek.

Time will show whether these efforts are engaged or whether Deepseek’s global rise will continue. Anyway, Amodei says he considers Deepseek a new competitor, which is at the level of the best AI companies in the United States.

“The new fact here is that there is a new competitor,” he said at Chinatalk. “In big companies that can train AI – anthrop, Openai, Google, maybe Meta and XAI – now Deepseek may be added to this category.”