ChatGPT Agent received high risk status for biological weapons
If you look at OpenAI’s ChatGPT Agent system card, you’ll find something surprising there. This AI system became the first in the world to officially receive “high risk” status for biological weapons development. Just think – technology that millions use daily is recognized as potentially dangerous. Link in description.
What does this status mean? According to official OpenAI documentation, Agent can significantly help even a non-specialist go through all necessary steps to create known biological or chemical threats. Until now, no company other than OpenAI has made such statements about their models.
Such recognition led to the introduction of unprecedented safety measures. Now tools for detecting potentially dangerous requests work not only at the response generation stage, but also before the request is even transmitted to the model.
Particularly interesting is that users are now required to report discovery of unusual system behavior. If you even accidentally encounter a strange model reaction, you must report it. Otherwise your account may be blocked. And you yourself may come under investigation.
