OpenAI has introduced a new safety architecture for the GPT-5.6 family to mitigate risks associated with its flagship Sol model. The system utilizes activation classifiers that monitor generations in real-time to intervene when sensitive domains are triggered. These safeguards are designed to block unsafe outputs across cybersecurity and...
Læs hele artiklen hos kilden.
Kommentarer (0)
Ingen kommentarer ennå. Bli den første til å kommentere!