Research / Sources
Research on reliable, interpretable and steerable AI. anthropic.com
Anthropic presents FlashProve, a zkSNARK-based system for cryptographic proofs of inference, letting providers prove responses came from a committed model, to support verification and compliance.
Anthropic's pilot risk report assesses misalignment risk from its deployed models as of summer 2025, concluding risk of catastrophic sabotage is very low but not negligible, under an affirmative-case framework.
Anthropic sketches hypothetical safety-case arguments for ASL-4 models with strategic and agentic capabilities, covering misuse, autonomous replication and autonomous AI research risks under its RSP.
Anthropic's Frontier Red Team discusses the difficulties of building robust, reliable model evaluations for safety and capabilities, and offers policy recommendations for evaluation-based AI governance.