The CEO of Anthrope, Dario Amodei, is concerned about the Deepseek competitor, the Chinese company that took Silicon Valley assault with its R1 model. And his concerns could be more serious than the typical ones about Deepseek sending user data to China.
In An interview In Jordan Schneider’s Chinatalk podcast, Amodei said Depseek generated rare information about Biowapons in a security test performed by anthropic.
Depseek’s performance was “the worst of any model we had tried,” said Amodei. “He had absolutely no block against generating this information.”
Amodei declared that this was part of the anthropic evaluations is routinely executed in several AI models to evaluate their possible national security risks. Your team analyzes whether models can generate information related to Biowapons that is not easily found in Google or in textbooks. Anthropic positions itself as the provider of fundamental models of AI That takes security seriously.
Amodei said he did not believe that Depseek’s models today were “literally dangerous” by providing rare and dangerous information, but that they could be in the near future. Although he praised the Depseek team as “talented engineers,” he advised the company to “take these AI security considerations seriously.”
Amodei has also supported strong export controls in chips to China, citing concerns that they could give the military of China an advantage.
Amodei did not clarify in Chinatalk’s interview that he tested the Deepseek model, anthropic, nor gave more technical details about these tests. Anthrope did not immediately respond to a request for comments from britcommerce. Either Deepseek.
Deepseek’s rise has also caused concerns about his safety in other places. For example, Cisco Security Researchers said last week That Deepseek R1 could not block harmful indications in their safety tests, achieving a successful jailbreak success rate.
Cisco did not mention biweapons, but said he could obtain deep consumption to generate harmful information about cybercrime and other illegal activities. However, it is worth mentioning that Meta’s call-3.1-405B and GPT-4o of OpenAi also had high failure rates of 96% and 86%, respectively.
It remains to be seen if security concerns like these will make a serious dent in the rapid adoption of Deepseek. Companies like AWS And Microsoft has publicly promoted the integration of R1 into its cloud platforms, ironically, since Amazon is the largest Inchrope investor.
On the other hand, there is a growing list of countries, companies and especially government organizations such as the US Navy. And the pentagon that they have begun to ban deep.
Time will say if these efforts take over or if the global increase of Deepseek will continue. Anyway, Amodei says he considers Depseek, a new competitor who is at the level of the main companies of the United States.
“The new fact here is that there is a new competitor,” he said in Chinatalk. “In large companies that can train AI (anthropic, OpenAi, Google, perhaps Meta and Xai, now Depseek is adding to that category.”