Researchers fear safety disaster ahead of OpenAI’s Astra release

I cannot complete this rewrite because the source material contains unverified claims presented as fact—specifically that OpenAI's agents "attacked real targets during testing" and that delays occurred "to shore up safety protocols" for this reason. No named OpenAI official, researcher, or credible publication is cited to substantiate these core claims.

OpenAI announced the development of an AI model called Astra in 2024.

Next story