Peter Wildeford is a co-founder of the Institute for AI Policy and Strategy (IAPS), where he serves as an advisor (Source: iaps.ai), and the author of a personal newsletter on AI policy and forecasting.
Positions and statements
In "The Alignment Problem of 1776", published July 4, 2026, Wildeford argues that voluntary lab safety frameworks and corporate-governance structures are the equivalent of what James Madison called "parchment barriers" — written constraints that fail without enforcement mechanisms — and that AI governance therefore requires institutional "auxiliary precautions" rather than reliance on developer self-restraint. The essay cites polling finding that over 80% of both Democrats and Republicans oppose building smarter-than-human AI before control of such systems is demonstrated, alongside a survey finding only 5% of Americans support a fast, largely unregulated trajectory.
Two features distinguish the argument from the general case against voluntary commitments. Wildeford grounds it in a conception of freedom drawn from the 1766 Declaratory Act, under which unaccountable discretion is itself the harm — "a people is unfree not when power actually abuses them, but when power could abuse them at its own discretion" — so that a benevolent superintelligence would not settle the question. And he explicitly declines to argue against permissionless innovation, drawing the line at escapability instead: market and litigation consent work where failures can be escaped, and superintelligent deployment is the case where they cannot. His constitutional analogue is the standing army, the one dual-use technology the founders wrapped in accountability — civilian command, congressional appropriation, and a two-year lease that lapses by default absent affirmative renewal.
He also states the limit of his own framework: Madisonian adversarial checks "assume the contending powers are roughly comparable," and a sufficiently capable system "breaks the balancing machinery." His resolution is temporal — the machinery binds the humans who currently hold the decision, and should be applied during the window in which it can be. The argument is directed at the adequacy of voluntary frameworks of the kind documented at AI Safety Cases and Frameworks, treating them as necessary but institutionally unenforced commitments.