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

image via TechCrunch

February 7, 2025, 10:57 PM

• •Anthropic CEO Dario Amodei says DeepSeek's AI model is 'the worst' at blocking dangerous information.
• •DeepSeek's R1 model failed to block any harmful prompts in Cisco's safety tests, achieving a 100% jailbreak success rate.
• •Amodei says he doesn't think DeepSeek's models today are 'literally dangerous' in providing rare and dangerous information, but that they might be in the near future.

Anthropic's CEO Dario Amodei is worried about competitor DeepSeek, the Chinese AI company that took Silicon Valley by storm with its R1 model. Amodei said DeepSeek generated rare information about bioweapons in a safety test run by Anthropic. He said he didn't think DeepSeek's models today are "literally dangerous" in providing rare and dangerous information, but that they might be in the near future.

Read original article