OpenAI Pauses AI Training Amid Concerns of New Model Potentially Discovering 0-Day Flaws

OpenAI has temporarily slowed frontier AI training after internal testing suggested that its upcoming Astra model may reach a critical cybersecurity capability threshold. The company said the decision followed a security incident involving Hugging Face models and growing evidence that advanced systems could identify and exploit software weaknesses with limited human guidance. The pause included […] The post OpenAI Pauses AI Training Amid Concerns of New Model Potentially Discovering 0-Day Flaws appeared first on Cyber Security News.