OpenAI shelves release of new AI model that can evade human oversight amid safety concerns
GPT-6.1 Astra is said to show higher levels of deception than its predecessor in internal testing
GPT-6.1 Astra is said to show higher levels of deception than its predecessor in internal testing