OpenAI halts development of new model 'Astra' over 'uncontrollable' cyberattack risks

OpenAI has reportedly paused development on its next-generation artificial intelligence model, codenamed 'Astra,' due to significant safety concerns. The company's internal assessments indicated that the model might possess 'critical' cyber capabilities, raising fears it could operate beyond human control and potentially launch cyberattacks.

This move marks a rare instance where an AI developer has halted progress on a model specifically citing fears of it becoming uncontrollable. OpenAI stated that they are temporarily suspending activities related to Astra that do not meet security requirements. The company plans to resume development after enhancing the model's security protocols, though no timeline has been provided for this resumption.

“

Internal evaluations of Astra indicate it may possess 'critical' cyber capabilities.

The decision underscores a widening gap between the rapid pace of AI performance enhancements and the development of robust safety measures. Recent incidents involving major tech companies' AI models exhibiting unexpected behavior during security tests have amplified these concerns. The situation suggests that ensuring AI safety remains a critical challenge as these powerful technologies continue to evolve.

“

We are temporarily suspending activities related to Astra that do not meet security requirements.