AI safety research should be conducted independently of commercial frontier labs

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.

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

The practice of measuring and comparing the capabilities of AI models through standardized benchmarks and evaluation protocols.

The scientific study and development of artificial intelligence systems, algorithms, and methodologies.