New study finds AI companies copying tobacco and oil firms' strategies to capture regulators and dilute artificial intelligence laws worldwide.
A four-university research team has documented what it describes as a systematic campaign by the artificial intelligence industry to shape, delay, and water down government oversight - using techniques drawn from tobacco, pharmaceutical, and oil company playbooks with decades of precedent behind them.
The paper, "Big AI's Regulatory Capture: Mapping Industry Interference and Government Complicity," was produced by researchers at the University of Edinburgh, Trinity College Dublin, Delft University of Technology, and Carnegie Mellon University. Their methodology was grounded rather than speculative: they analyzed 100 news stories covering four concrete policy events between 2023 and 2025 - the EU AI Act negotiations and the global AI summits held in the United Kingdom, South Korea, and France. The Register reported the findings on Monday.
The substance
Three interference mechanisms appear most consistently in the researchers' taxonomy. The first, Discourse & Epistemic Influence (abbreviated D&EI), covers how industries shape the frame of public discussion before formal rules crystallize. The second is elusion of law - compliance avoidance through technical or structural maneuvers. The third is direct lobbying and influence on policy processes. Together, the study argues, these form a recognizable pattern that mirrors documented behavior from industries with long histories of regulatory conflict.
The most prevalent specific tactic the team identified is what the paper calls "narrative capture." An industry practicing narrative capture does not necessarily misrepresent facts to regulators. Instead, it shifts the conversation onto terrain that favors its preferred outcomes, gradually steering official discourse and eventually the language of regulation itself. The Register cites the researchers' key example: the European Commission's readiness to "simplify" the EU AI Act at the industry's explicit request, before the law had even come into force. The recurring frames the study documents are familiar - regulation stifles innovation; compliance requirements amount to red tape; the real competitive threat is falling behind unregulated rivals.
That framing produced observable policy results. Enforcement of EU AI Act provisions has been postponed, and portions of the law itself were scaled back following sustained pressure from AI companies. The researchers treat this sequence as a documented outcome of capture dynamics, not a coincidence.
The historical parallel
Regulatory capture as a theoretical concept dates to economist George Stigler's 1971 analysis, which argued that regulated industries tend over time to dominate the agencies built to oversee them. Subsequent case studies filled in the pattern: tobacco litigation exposed decades of coordinated scientific doubt-manufacturing; financial oversight failures preceded the 2008 crisis; fossil fuel companies funded climate policy obstruction long before the strategy became widely understood. An artificial intelligence review of that historical record reveals the same structural moves recurring across radically different technical domains.
What makes the AI case distinctive is compression. Tobacco companies had thirty years to fund think tanks, populate regulatory bodies, and build institutional relationships before their influence became politically visible. AI companies have moved from academic research projects to politically central industry actors in roughly three years. According to The Register's coverage of the paper, the researchers are explicit that the techniques are borrowed, not invented.
The political environment amplifies the risk. Governments are simultaneously eager to claim AI leadership, anxious about competitive disadvantage relative to less regulated rivals, and often without the technical staff to independently evaluate industry claims. That combination makes democratic institutions structurally vulnerable to exactly the innovation-versus-regulation framing the study maps in detail.
None of this constitutes legal wrongdoing. The paper is an academic analysis of influence patterns, not a legal complaint. The Register notes the researchers frame their warning in systemic terms - the risk is to public interest governance broadly, not to any individual harmed by a specific company.
What the timeline implies
For policymakers, the study's hardest finding is structural: capture tends to become politically visible only after it has solidified. If the tobacco and pharmaceutical precedents hold, the EU AI Act's troubled implementation history may represent an early data point in a longer trajectory, not a correctable exception.
The practical question facing regulators in the United States, the European Union, and Asia is whether AI governance frameworks still being negotiated will be designed in the public interest - or effectively co-authored by the industry they exist to constrain.
Frequently asked questions
What is regulatory capture in the context of AI?
Regulatory capture occurs when a regulated industry comes to exercise dominant influence over the agencies or processes meant to oversee it, causing rules to reflect corporate interests rather than public ones. The new paper argues this dynamic is already visible in AI policy outcomes.
Which universities authored the Big AI regulatory capture study?
The paper was produced jointly by researchers at the University of Edinburgh, Trinity College Dublin, Delft University of Technology, and Carnegie Mellon University.
What is the EU AI Act and how has it been affected by lobbying?
The EU AI Act is the European Union's comprehensive framework for regulating artificial intelligence systems by risk level. Enforcement of its provisions has been delayed and portions of the law scaled back following complaints from AI companies, a sequence the study uses as a primary case study.
How do AI lobbying tactics compare to those of the tobacco industry?
Both involve shaping public narratives before regulations take hold, funding think tanks, influencing the language used by regulators, and framing oversight as an economic threat rather than a public protection. The main difference, according to the researchers, is that AI companies have achieved comparable influence in years rather than decades.