OpenAI blog: Safety and alignment for long-horizon models

The post examines safety and alignment challenges for AI models that operate over extended time horizons. It discusses potential risks and mitigation strategies for long-horizon decision-making.