OpenAI has identified its model Astra as the first to meet the ‘Critical’ cybersecurity capability threshold in the company’s Preparedness Framework, a designation that will trigger stronger protective measures before the model is released more broadly.
According to OpenAI, the Preparedness Framework links defined capability thresholds to the application of safeguards. Astra’s classification as ‘Critical’ signals that the model has attained a level of capability that requires additional security precautions prior to wider availability.
OpenAI said it will apply heightened safeguards to Astra as part of aligning deployment decisions with measured safety and security controls. The company framed the step as an element of its broader approach to matching model releases with appropriate protective measures based on assessed capabilities.
The announcement did not include further technical details about Astra or a timeline for its release. OpenAI’s statement emphasized the continued practice of tying capability assessments to protective actions as more advanced models are developed and evaluated.
Astra’s designation under the Preparedness Framework represents a notable milestone in OpenAI’s internal processes for evaluating and managing cybersecurity risks associated with increasingly capable AI systems. For additional information, OpenAI refers readers to its original post.
Source: Read the original source
Leave a Reply