When will a catastrophic AI duty of care become U.S. law?

On the substance, the case for Yes is stronger than the market price implies if the contract resolves on any U.S. law rather than exclusively federal law. California’s SB 53 has already imposed frontier-developer transparency and incident-reporting obligations, New York’s RAISE Act is moving into effect around the deadline, and Colorado has reenacted an AI framework with additional duties and implementing rules. Taken together, these developments show that the U.S. is not waiting for Congress to create AI safety duties; states are already building a legal regime that can credibly be described as a catastrophic-risk duty of care or at least a close functional equivalent.

The main counterargument is definitional and procedural. A true “duty of care” usually implies something more than reporting, documentation, or transparency, and some of the cited state statutes may not cleanly satisfy a strict reading of catastrophic-risk duty of care. If the market contract expects a federal enactment, the outlook is much weaker because Congress remains stalled and there is still no broad national AI safety law on the books. That distinction matters a lot, because a federal-only reading would make No far more likely than Yes.

Compared with the current market price of 8%, the market looks too dismissive of the possibility that state law counts and that the relevant statutes are already in force or imminent before January 1, 2027. The market is probably anchoring on the federal bottleneck and underweighting resolution ambiguity. My estimate is still well below a coin flip only because there is real risk that the contract will be interpreted narrowly and exclude state-level obligations or reject reporting duties as insufficiently close to a duty of care.