Can big tobacco explain big tech?

Artificial intelligence is forcing governments to confront a familiar regulatory challenge: how should societies govern powerful industries before their products become so deeply embedded that meaningful oversight becomes difficult?

One historical comparison has attracted particular attention. Some policymakers and technology critics argue the tobacco industry's decades-long campaign to dispute scientific evidence, lobby policymakers and promote voluntary self-regulation offers lessons for governing today's largest technology companies. The response eventually helped shape the WHO Framework Convention on Tobacco Control, adopted in 2003 and now ratified by 183 Parties, which established international standards on tobacco advertising, disclosure, taxation and limits on industry influence over public health policy.

Nell Watson, AI ethicist and president of the European Responsible Artificial Intelligence Office, said the comparison is compelling because it captures the incentives facing powerful companies, but becomes less persuasive when applied to artificial intelligence itself.

The tobacco frame certainly captures the corporate incentive problem well," Watson told The Beiruter. "But it may misdescribe the technology’s impacts.

The question facing policymakers is therefore larger than whether AI resembles tobacco. It is whether governments can apply history's lessons before regulation falls behind innovation.

The corporate playbook

The comparison between Big Tech and Big Tobacco is less about the products themselves than how powerful industries respond when growing evidence of harm collides with commercial incentives.

That comparison stems largely from the tobacco industry's response to mounting scientific evidence linking smoking to disease. Rather than accepting growing evidence of harm, tobacco companies spent decades funding research that cast doubt on scientific findings and lobbying governments to delay regulation. Internal documents later made public through litigation revealed that executives often understood far more about smoking's health risks than they disclosed.

Watson sees echoes of that pattern in parts of the technology sector.

"The internal-documents pattern is a familiar one," she said.

Companies often tacitly know more about their products' harms than they disclose, and we tend to learn the details through leaks and whistleblowers rather than voluntary transparency.

In Watson's view, the similarities extend beyond transparency. She argues some technology companies have responded to concerns about AI by emphasizing uncertainty around long-term risks and framing smoking as a matter of personal rather than corporate responsibility. She also points to business models that reward companies for keeping users engaged for as long as possible, even when that engagement can encourage compulsive behavior and shift many of the resulting social costs onto users and society.

That does not mean today's AI companies are equivalent to tobacco companies, nor that artificial intelligence presents the same kind of risks. Rather, Watson argues that the comparison is useful because it highlights how profitable industries often respond when asked to regulate themselves before the consequences of their products are fully understood.

What tobacco regulation actually teaches

The history of tobacco regulation offers lessons that go well beyond the dangers of smoking.

Adopted in 2003, the WHO Framework Convention on Tobacco Control became the world's first public health treaty negotiated under the World Health Organization. Rather than prohibiting tobacco, it established an international framework built around advertising restrictions, health warnings, higher tobacco taxes, action against illicit trade and protections against secondhand smoke.

Article 5.3 may be its most enduring contribution. It requires governments to protect public health policy from the tobacco industry's commercial interests after decades of evidence that lobbying had delayed stronger regulation.

The WHO's 2025 Global Tobacco Epidemic Report suggests this gradual approach has had measurable effects. More countries have adopted comprehensive smoke-free laws, graphic warning labels and advertising restrictions than at any point in the treaty's history.

Where the analogy breaks down

The analogy begins to weaken, however, once attention shifts from corporate behavior to the technology itself.

"Tobacco is a product with little apparent redeeming societal value," Watson said.

AI, on the other hand, is a general-purpose technology with enormous legitimate benefits in medicine, science, accessibility, and productivity.

That distinction, she argues, is fundamental. Unlike cigarettes, whose health effects became measurable through decades of epidemiological research, AI does not produce a single, universally recognized form of harm. Its risks range from misinformation and algorithmic bias to privacy concerns, labor displacement and potential misuse, many of which remain contested or may only emerge over time. As a result, AI presents a far more complex governance challenge than regulating a product whose primary harms eventually became well established.

RAND, a U.S.-based nonprofit research organization, reaches a similar conclusion in its 2024 report Historical Analogues That Can Inform AI Governance. Rather than endorsing any single historical comparison, it argues that different technologies illuminate different governance challenges. Nuclear power offers lessons on international oversight, aviation on safety institutions and biotechnology on scientific governance. Policymakers, it concludes, should draw selectively from multiple historical experiences rather than searching for one perfect analogy.

A better lesson from history

While Watson believes the tobacco analogy helps explain corporate behavior, she argues it is a less useful comparison for artificial intelligence itself.

“Another, potentially stronger analogy is fossil fuels,” she said.

Like AI, hydrocarbons delivered transformative and real benefits that made whole economies dependent on them, while generating externalities that were systemic, delayed, and unevenly distributed.

Watson argues the lesson from fossil fuels is not prohibition but acting before technological dependence makes course correction significantly more difficult.

That philosophy is reflected in today's leading AI governance frameworks. The OECD AI Principles emphasize transparency, accountability and human oversight, while NIST's AI Risk Management Framework treats governance as an ongoing process of identifying, assessing and managing risks as systems evolve.

Watson sees aviation as a useful model for how that oversight should evolve. Commercial aviation became dramatically safer not through banning flight after accidents occurred, but through independent accident investigations, mandatory incident reporting and a regulatory culture built around continual learning and safety improvements. Watson argues AI governance should follow a similar path, creating institutions that identify failures early and adapt alongside the technology.

History, then, suggests the challenge is not deciding whether artificial intelligence most closely resembles tobacco, fossil fuels or aviation. It is building institutions capable of measuring risk independently, responding to new evidence and ensuring innovation advances alongside accountability.