Anthropic's statehouse AI doctrine is a moat tell

When Washington cannot act, the regulatory posture a company pushes in fifty statehouses is a competitive tell: the doctrine it favors reveals which moat — which durable market advantage — it is trying to harden into law. Anthropic is lobbying for state laws that keep ratcheting safety requirements upward for the most capable AI systems, meaning the largest, most powerful models rather than chatbots. OpenAI, by contrast, is pushing to copy the same bill from state to state, a strategy its top lobbyist Chris Lehane calls "reverse federalism" — state legislatures mirror one chamber's bill until it effectively becomes a national floor.

Both companies claim to be competing to protect the public. The pattern says something else. Anthropic's ratchet raises the compliance surface, the legal and engineering cost any rival must absorb, and defines "most capable" on thresholds Anthropic-shaped labs already meet. OpenAI's mirror doctrine flattens that surface and protects a scale and developer-ecosystem moat, the advantage that comes from shipping the most product, supported by the most third-party apps. Trump-adjacent and venture-capital critics call the Anthropic line a lockout: a safety rule written for the company that already clears it. Wire coverage will frame this as Anthropic versus OpenAI on AI rules. The mechanism is older: when the federal floor is missing, every statehouse bill is a competitive move, and the doctrine a company lobbies for is the moat it wants lawmakers to ratify.