OpenAI Tightens Security Controls on Astra Model

OpenAI has added new limits and security safeguards to its Astra model. The company says Astra may be a cybersecurity risk. OpenAI is limiting the model’s access to cybersecurity capabilities.

Astra was trained to refuse harmful requests and spot jailbreak attempts. OpenAI said Astra was not the model involved in the Hugging Face breach.