OpenAI pauses development of new model Astra over concerns it may possess critical cyberattack capabilities

Odaily News: OpenAI stated that it cannot rule out the possibility that its upcoming model, Astra, possesses "critical" cyber capabilities, prompting OpenAI to expand its safety testing scope and suspend internal activities that do not meet stricter security requirements. This is the latest sign of the rapid advancement of AI models' cyber capabilities.

OpenAI said that after an internal assessment of one of its upcoming models, Astra, "we cannot rule out that it possesses critical cyber capabilities." Ahead of Astra's release, OpenAI will expand testing and safety measures for the model and, in accordance with the "Preparedness Framework" first published by the company in 2023, slow down Astra's development process until appropriate safety safeguards are established.

OpenAI also stated that Astra was not involved in the Hugging Face vulnerability attack incident. This may be the first case of an AI frontier lab committing to slowing down its own AI model development due to cybersecurity concerns. (Axios)