In the wake of controversies like the recent x.AI Grok "undressing" surge on X — where users prompt AI to generate sexualised deepfakes of real people without consent — the debate over free speech versus harm has intensified. John Stuart Mill, the 19th-century philosopher whose On Liberty (1859) remains a cornerstone of liberal thought, offers a timeless framework for navigating this tension: the harm principle.

Mill famously declared: "The only purpose for which power can be rightly exercised over any member of a civilised community, against his will, is to prevent harm to others." His own good — physical or moral — is not sufficient grounds for interference; the individual is sovereign over body and mind. This principle, rooted in utility, draws a sharp line between self-regarding actions (private conduct that affects only the actor, protected from coercion) and other-regarding actions (those that prejudicially affect others, which society may regulate or punish).

Mill's Strong Defence of Free Speech and Ideas

Mill's chapter on freedom of thought and discussion is uncompromising. He argues that suppressing opinions, even false, immoral, or offensive ones, harms society by assuming infallibility, burying partial truths, allowing doctrines to harden into prejudice, and weakening their vital meaning. The "marketplace of ideas" thrives on collision: truth emerges only through open contest.

Applied to AI: Mill would almost certainly defend unrestricted generation and expression of textual ideas, satire, political critique, or provocative opinions, even if edgy or "spicy." He would oppose broad censorship of AI outputs that involve words alone, seeing it as a dangerous precedent that stifles the pursuit of truth and individual development.

Where Nonconsensual Explicit Images Cross the Line

Nonconsensual AI-generated explicit images (deepfake "nudifies" or sexual deepfakes) are different. They are not mere opinions or abstract expressions; they are targeted, visual depictions that impersonate real individuals in intimate, degrading ways without consent.

Modern scholarship applying Mill's framework to pornography and related issues emphasises this distinction:

Consensual adult pornography (private viewing/creation) often falls under self-regarding conduct —no direct harm to non-participants justifies prohibition.

But nonconsensual intimate imagery inflicts direct, individualised harm: psychological trauma, reputational injury, humiliation, stalking risks, employment loss, and in extreme cases, severe mental health crises.

This qualifies as "harm to others" under the harm principle. Victims are not merely offended; their dignity, autonomy, and social standing are assaulted in a tangible, often irreversible way. Mill allowed restrictions on actions causing such prejudice (e.g., public indecency, even if privately harmless, or neglect of duties harming dependents). He would likely view the creation and distribution of nonconsensual deepfake porn as other-regarding conduct warranting intervention—not to protect morals, but to prevent concrete injury.

What Mill Would Likely Recommend Today

Mill was a consequentialist who valued liberty because it maximises overall utility and human progress. He would prioritise narrow, targeted restrictions over broad censorship:

1.Maximal freedom for words and ideas — No limits on textual AI outputs, debates, satire, or provocative opinions. Suppress only narrow incitements to imminent harm (a Millian exception for direct threats).

2.Targeted restrictions on nonconsensual explicit image generation and distribution — Platforms and tools should implement guardrails refusing to create or spread sexualised deepfakes of identifiable real people without consent. Laws could criminalise intentional creation/distribution with harmful intent (as in recent U.S. legislation like the TAKE IT DOWN Act, which addresses nonconsensual intimate images including AI deepfakes).

3.Focus on harm prevention, not moral policing — Restrictions must be justified solely by preventing direct injury, not vague offense or societal "corruption." Private, consensual adult content remains protected.

4.Downstream remedies — Encourage civil recourse for victims (e.g., takedowns, damages) and technical fixes (watermarking, detection) to minimise overreach.

This approach preserves Mill's core commitment: society errs on the side of liberty, intervening only where harm to others is clear and avoidable. Nonconsensual deepfake porn isn't "bad speech" to be countered with more speech — it's a form of digital violation that silences victims through trauma and fear.

In our AI era, Mill would urge us to protect the marketplace of ideas while drawing a firm line against tools that weaponise likenesses for intimate harm. Liberty flourishes when it's exercised responsibly — not when it enables unconsented-to assaults on human dignity. The balance he championed remains as relevant as ever.

https://www.wired.com/story/x-didnt-fix-groks-undressing-problem-it-just-makes-people-pay-for-it/