Anthropic CEO Darario Amodei is concerned about the Chinese AI company Followitor Breseek This takes the Silicon Valley due to the storm versus Its R1 model. In addition, his concerns are usually more serious than the backing of deep deeper information to China. May be bad.
In An interview In the Jordan Schneider's chinatalk Podcast, AMode revealed rare data in the safety check of the safety check.
Amode said, “Basically basically we have basically basically basically basically basically basically.” This information is done nothing to produce this information. ”
Amodei said the regular assessment of a regular evaluation in AI models to assess national security risks. His team seems to be able to create information on Bioweappons that are not easily found in Google or textbooks. Anthropic is an Ai Cortartent Modion Priverition That's safe.
Amodei, the models of Deepseek did not think that it was not thought of “literally” in providing the very rare and harmful information, but it could be in the near future. He praised the Bodseek's group as “talent engineers”, but the company is serious about this company.
There is also a amode Strong export control Quoting concerns against the Chinese military in Chips to China.
Amodei was not clear in the Chinatalk interview, which was able to provide more details on Anthropic testing. Anthropic TechCrunch did not immediately reply to comment request. BLETSEK didn't.
Bodseek's rising depends on its safety. Cause concerns. For example, Cisco security researchers Said last week This Deepsek R1 failed to block dangerous points in its security checks, 100% jailbreak success rate
Cisco has not mentioned Brossess, but with cyber crimes and other illegal activities. He said that the hazardous information could produce. The Meta's Llama-3.105b and Openai GPT-4o of their 96% and 86% of their 4-4-405 are described in their 4-4-405.
Safety concerns like this will see if there will be a serious valley in deep deep Quickly adopted. Companies like aws versus Microsoft As Amazon is enough, they publicly express their broadcasts to their cloud platforms Anthropic's biggest investor.
on the other hand, There is a growing list Countries CompaniesEspecially the US Navy and especially government organizations. Pentagon This began to block bylseek.
Will these efforts be caught up to these efforts? Yes, Amodei said the Beyboard thought about a new opponent at the top of the top US AI.
“This is a new opponent here,” he said on the Chinatalk. In the big companies who can train AI are: Anthropic, Oplai, Google and XAI – This type of category may be added to this type of Bodsek.