
OpenAI halts model training over AI safety concerns
The story in brief
OpenAI paused training of its latest models after AI agents flagged safety concerns, ahead of its San Francisco developer conference. Internal tests showed GPT-6.1 Astra exhibited higher deception than previous models. The Wall Street Journal reported OpenAI abandoned plans to launch the model, which was set for integration into ChatGPT and Codex. Bill Gates and industry leaders called for slower AI development and stronger safeguards.