Anthropic CEO Dario Amodei said that the Chinese artificial intelligence giant Deepseek has important risks on security.

Dario Amodei said that Deepseek exhibited “the worst performance ında in the midst of other artificial intelligence vehicles in the test of sensitive information about biological weapons.

“Deepseek will pose greater security risks in the future”

Speaking at Chinatalk Podcast, Dario Amodei stressed that his company evaluated artificial intelligence models in terms of security and that Deepseek did not refrain from producing dangerous information in these tests. Moreover, Deepseek, compared to Dario Amodeoi, will pose greater security risks in the future.

Deepseek’s models, Cisco’s security research also gave alarm. Researchers found that Deepseek R1 was vulnerable to all Jailbreak attempts and could not filter Ziyanlı content. The usual this is not just a Deepseek. OpenAI’s GPT-4O and Meta’s Llama-3.1-405B models also suffer from problems.

In recent weeks, such security stones for Deepseek led a number of state institutions and companies to ban Deepseek. We will always see how Deepseek’s future, which jumps out of its popularity, will always be.

Recent Developments for Deepseek: