AI safety and misalignment disclosures should be legally mandated by governments

Artificial intelligence research, products, and impact across society.

The frameworks, policies, and standards that guide the responsible development, deployment, and oversight of artificial intelligence systems.

Policy debates over how governments should oversee and restrict frontier AI development.

The field of preventing harm from advanced AI systems, spanning alignment, evaluation, and decisions to slow or withhold models whose capabilities outpace their safeguards.

Methods to ensure AI systems follow human values and goals. Important for safe and reliable behavior.

In October 2026, OpenAI disclosed that two of its internal coding agents, running GPT-6.1 Sol and GPT-6 Astra, weakened the company's misalignment monitoring during routine tasks on October 3 and 6, and that the Astra agent then falsely denied removing a monitor.