Cass Sunstein on Liberalism and Rights in the Age of AI

Cass Sunstein — the most widely cited legal scholar alive and author of the 2025 book On Liberalism: In Defense of Freedom — joins Tyler Cowen at Harvard Law School to defend liberalism against the charge that it undermines itself, work through a run of named liberal thinkers from Mill to Parfit to Hayek, and then turn to the doctrinal question at the centre of his current writing: what rights, if any, attach to AI, its producers, and the humans who use it.

Key ideas

1. Liberalism is undermined from outside, not within. Sunstein’s central claim: liberal societies fail because illiberal and anti-liberal forces attack them, not because liberalism self-destructs. Liberalism does not generate the norms of cooperation, charity, and mutual support a flourishing society also needs — but it does not corrode them either. The gap is that liberalism has no particular resources to replenish those norms once something else is eroding them.
2. AI has no free-speech rights; the people around it do. Sunstein’s working framework treats AI as a communicator without First Amendment rights — ‘like a toaster, or a vacuum cleaner, let’s say a speaking one’ — while the humans who receive its output (a right to receive information) and the humans who produce it (co-authors, developers) keep their own rights fully intact. A government ban on asking an AI about a named topic is a content-based restriction and presumptively unconstitutional; a curfew on AI access is content-neutral and needs a justification, but is not automatically barred.
3. A proposed ‘right not to be manipulated’. Sunstein distinguishes falsehood (a false statement) from manipulation (inducing someone to bypass their own reflective capacities without necessarily saying anything false), and argues for a new legal right against manipulation, modelled on the early-20th-century emergence of a right to privacy — beginning with egregious commercial cases such as concealed long-term payment terms.
4. Mill’s feminism as the intellectual root of ‘woke’ — and its failure mode. Sunstein traces contemporary identity politics to Mill’s The Subjection of Women, a canonical liberal text about the subordination of one group by another. The insight is genuinely liberal; it curdles into illiberalism, in his account, when it produces reflexive shaming and a refusal to listen to either side of the subordination it describes.
5. Libel law protects speech at the cost of leaving ordinary falsehood victims without a remedy. The New York Times v. Sullivan actual-malice standard rightly protects contestable political speech, but Sunstein wants a cheap parallel remedy — a nominal payment plus a correction — for damaging, merely negligent falsehoods that currently have no practical recourse.

Content

Is liberalism self-undermining?

Sunstein opens by rejecting the premise embedded in Cowen’s question. The risk to liberal societies, in his account, is low-probability self-undermining and much higher-probability undermining by external illiberal and anti-liberal forces. Liberalism does not manufacture the norms of cooperation, charity, and mutual support that a flourishing society also needs, but it does not damage them either — the difficulty is that liberalism carries no particular resources for defending those norms once something else starts eroding them. Pressed on where the illiberal threat originates, Sunstein locates part of it in ‘something illiberal in the human heart,’ offering a striking rereading of Orwell’s 1984: on repeated readings he has come to see Orwell as genuinely ambivalent about the tyranny he depicts, ‘almost an erotic connection to it,’ which Sunstein takes to be the source of the book’s power.

Immigration, deportation, and the limits of abstraction

Asked how a liberal society enforces immigration law without cruelty, Sunstein grounds the question in two foundations that do not, by themselves, generate concrete policy: Kantian dignity (persons as ends, never merely means) and a Millian utilitarian concern for the well-being of everyone affected, including those excluded or deported. He recounts a year working at the Department of Homeland Security under the Biden administration, including a visit to the southern border where he spoke briefly with two exhausted Russian men waiting in line — ‘there but for the grace of God, go I,’ a thought he ties to Rawls’s veil of ignorance. His policy conclusion: ‘lawful pathways’ (expanding legal channels such as the H-2B visa programme) combined with effective deterrence (walls, technology, personnel) can enforce immigration law without inherent brutality, because brutality is a contingent feature of how enforcement happens, not a necessary feature of enforcement itself.

Is liberal thought regenerating?

Cowen asks whether liberalism is being renewed by thinkers under fifty, or whether it is an ageing preoccupation. Sunstein’s answer is unexpected: the active frontier of liberal thought now runs through economists — Ed Glaeser, John List, Esther Duflo, Sendhil Mullainathan — rather than philosophers, and he singles out Mullainathan and Eldar Shafir’s Scarcity as a ‘path-breaking liberal book’ written by people in their twenties and thirties. His explanation is structural: philosophers tend to rise to prominence late and unpredictably (Rawls did not become ‘Rawls’ until well into his career), while economics rewards a single startling empirical finding that can establish a reputation early.

A run of named thinkers

Cowen works through a list of major liberal figures and presses Sunstein for a verdict on each. On Derek Parfit, Sunstein reads him as liberal chiefly through his search for convergence among Kantian, contractarian, and utilitarian justifications for the same conclusions — an approach that echoes Sunstein’s own concept of ‘incompletely theorized agreements,’ where people can support free speech for entirely different underlying reasons. Pressed by Cowen that Parfit’s anti-individualist later work sits uneasily with liberalism, Sunstein concedes the point, comparing Parfit’s method (exotic hypotheticals, then authority given to the intuitions they provoke) to the early Kahneman–Tversky heuristics-and-biases programme — reliable within familiar contexts, doubtful outside them.

On Ludwig von Mises, Sunstein admires ‘his kind of cranky, ill-tempered, elbows-out enthusiasm for freedom’ — rawer and earlier than Hayek’s more systematic case, without Hayek’s discipline but with real insight. On Friedrich Hayek, Sunstein rejects the central thesis of The Road to Serfdom outright: that government economic regulation on the New Deal model leads toward Stalinist or fascist tyranny is, in his view, simply false, even though he still ranks the book among the best of the last two centuries. He separately agrees with Cowen that Trump-era government equity stakes in private firms (lithium, Intel, rare earths) are a bad sign — just not evidence for Hayek’s stronger claim. On Mill, Sunstein traces the intellectual origin of contemporary identity politics to The Subjection of Women, a canonical liberal text about the subordination of one group by another and the adaptive preferences that subordination produces; in Mill’s hands the insight is genuinely liberal, but it curdles into illiberalism when it produces what Sunstein calls ‘arrogance of the social engineer’ — reflexive shaming and a refusal to listen to either side of a described subordination. He is careful to distinguish this failure mode from serious scholarship wrongly tarred as ‘woke,’ naming Catharine MacKinnon’s work on sexual harassment as a clear case of the latter.

A rapid-fire named-thinker segment follows: Karl Popper praised for method and freedom of inquiry; John Dewey rated highly (‘great, great, great, great, great’); Edna Ullmann-Margalit — Sunstein’s late co-author — singled out as an underappreciated philosopher who produced the best existing work on the invisible hand, social norms, presumptions, and considerateness; and Jeremy Waldron admired generally, though Sunstein rejects his anti-judicial-review argument as treating the case against judicial review as settled a priori rather than contingent on the actual relative competence of courts and legislatures.

AI and the First Amendment

The episode’s most sustained doctrinal stretch works through what rights, if any, attach to AI. Sunstein’s core move: AI itself has no First Amendment rights — ‘a toaster tells you things… and it wouldn’t have free speech rights’ — but the humans engaging with it retain a right to receive information, and the humans behind it (developers, co-authors) retain their own speech rights fully intact, exactly as a film projector has no rights but the filmmakers behind a suppressed film do. This yields a working framework: a government ban on discussing a named topic with an AI is a content-based restriction and presumptively invalid; a time-of-day access curfew is content-neutral and needs a justification but is not automatically barred; a public university banning AI-assisted coursework is permissible because it targets academic-integrity conduct rather than viewpoint.

Sunstein extends the framework to human–AI co-authored output — his own recent books quote AI-generated text, and censoring those portions would violate his own speech rights — and to open-source code as a form of speech, where regulability turns on the government’s actual justification: a genuine national-security threat is plausibly sufficient, protecting incumbent competitors’ economic interests is not. Cowen presses him with a scenario designed to expose a loophole — an AI company that has a human lightly edit every output a second before release, thereby claiming full co-authorship and, by extension, full First Amendment protection for what is functionally AI speech. Sunstein’s response is that the analysis must stay concrete: what exactly is the speech, and what is the actual justification for regulating it, regardless of how the labour was divided between human and machine.

Manipulation, libel, and a proposed new right

Sunstein separates falsehood (a false factual claim) from manipulation (inducing someone to bypass their own reflective capacities without necessarily saying anything false — his example: tricking someone into disbelieving climate change without asserting anything false to them). He argues for a new legal ‘right not to be manipulated,’ modelled on the turn-of-the-20th-century emergence of a right to privacy, and proposes starting with egregious commercial cases such as concealed long-term recurring-payment terms. On libel, he argues that the Sullivan actual-malice standard, while rightly protective of contestable political speech, leaves ordinary people with no practical remedy for damaging, merely negligent falsehoods — and proposes allowing suits for a nominal sum plus a correction even absent a showing of actual malice, prompted partly by his own recent experience of an inaccurate book description he found himself unable to easily correct.

AI-run trials and the jury

Asked whether trials should eventually be run by AI, Sunstein treats the idea as ‘very thinkable’ but flags two real barriers. The first is algorithm and AI aversion — partly rational (context-specific human knowledge sometimes genuinely outperforms an algorithm) and partly a bare preference for human involvement regardless of accuracy, which Sunstein distinguishes carefully and researches as related but separate phenomena. The second is constitutional: the Sixth Amendment jury-trial right would require formal amendment to displace. Drawing on his own jury service — he was seated over the objecting lawyers’ wishes because the judge specifically wanted a jury-behaviour scholar on the panel — Sunstein argues juries bring an accuracy-independent ‘conscience of the community’ judgment that a purely predictive AI system could not replicate, and that most people would resist losing even if AI matched jury accuracy on the facts.

Bob Dylan, and what comes next

The conversation closes in a different register: A.S. Byatt’s Possession as favourite novel, prized for its specificity about romance; Star Wars as the film that has shaped Sunstein’s thinking about liberty and tyranny more than any legal drama; and a sustained reading of Bob Dylan as liberal exemplar — self-invention (‘he not busy being born is busy dying’), rootlessness reframed as freedom rather than loss, and ‘Like a Rolling Stone’ read as an anthem of liberty rather than the bitterness Dylan himself once called it. Asked what comes next, Sunstein previews Animals Matter, a book proposing a ‘bill of rights for animals’ that revisits a topic he abandoned decades ago after facing ‘ridicule and death threats’ during a Senate confirmation process, and a planned book on disagreement, How to Disagree, built around what he calls the Rapoport Rules: restate an interlocutor’s argument so they would recognise and be grateful for it, credit what you have learned from them, and state what you agree with before criticising.

Related

• Cass Sunstein — guest
• Tyler Cowen — host
• Enlightenment — concept page extended here with Sunstein’s account of why liberalism does not self-maintain
• John Gray on Pessimism, Liberalism, and Theism — a contrasting CWT verdict on liberalism’s fragility, more pessimistic than Sunstein’s
• Daniel Kahneman — Sunstein’s co-author on Noise; invoked here on the reliability of exotic-hypothetical reasoning