Topic: OpenAI Astra

OpenAI is sitting on evidence that its next major AI model could cross a threshold the company itself defined as "Critical" for cybersecurity risk. The catch: they won't release the data to prove it, or disprove it. That tension is drawing serious scrutiny from AI safety researchers and independent evaluators. For anyone paying attention to where this technology is heading, the details matter.