OpenAI Previews Astra: AI Cybersecurity Model Reaches 'Critical' Preparedness Threshold

OpenAI ·

Key Info

OpenAI is preparing to release Astra, an AI model for cybersecurity that has reached the "Critical" threshold under its Preparedness Framework. The company is previewing its evaluation methods, the safeguards built alongside the model's capabilities, and areas for continued improvement.

Highlights

  • Astra marks a significant advance in cybersecurity capability, meeting the highest risk tier in OpenAI's internal safety framework.
  • OpenAI is sharing how the model was evaluated and how safety measures advanced in parallel with capability.
  • The release will be accompanied by ongoing learning and improvements, reinforcing a focus on safe and broad accessibility.
Loading...