Philip Morris used AI to generate public opposition to EU rules

2026-06-29 · Source: Welcome to the Artificial Intelligence Incident Database · Field: Government & Public Sector — Public Policy & Governance, Regulatory & Compliance, AI in Public Affairs · Depth: Fundamental Awareness, quick

Summary

Tobacco giant Philip Morris International (PMI) employed AI to create hundreds of fabricated "citizen" responses opposing stricter European Union regulations on vaping products. An investigation by broadcaster NOS and journalism platform Pointer, utilizing the AI-detection tool Pangram, found that approximately 30% of 65,000 analyzed responses from 15 countries were AI-generated, with the Netherlands showing a 71% rate. PMI's campaign involved placing QR codes in tobacco shops, directing users to a tool that generated first-person texts for submission to the European Commission. This effort resulted in 80,000 total responses, significantly higher than typical consultations. The European Commission expressed "very concerned" about these "alleged attempts," while Dutch MPs condemned the actions as "cheating" and "hijacking democracy." PMI defended its use of AI for readability, calling it a "legitimate contribution."

Key takeaway

For policy makers evaluating public consultations, this incident highlights the critical need to verify the authenticity of submitted feedback. Philip Morris's use of AI to generate responses demonstrates how sophisticated tools can distort democratic processes and create artificial grassroots support. You should implement advanced AI detection mechanisms and update consultation guidelines to explicitly address AI-generated content. This proactive approach is essential to maintain the integrity of public input and ensure regulatory decisions reflect genuine citizen sentiment.

Key insights

Philip Morris used AI to generate fake public comments, simulating grassroots opposition to EU vape regulations.

Principles

• AI can simulate grassroots public opinion.
• Unusual consultation response volumes warrant scrutiny.
• AI detection tools can expose synthetic text.

Method

Philip Morris deployed QR codes in shops, linking to a tool that generated first-person texts from user input for submission to the European Commission's public consultation.

In practice

• Employ AI detection for public feedback.
• Monitor consultation response volume.
• Scrutinize templated text campaigns.

Topics

• AI Lobbying
• Public Consultation
• Tobacco Regulation
• AI Detection
• Democratic Integrity
• Misinformation

Best for: CTO, VP of Engineering/Data, Director of AI/ML, Policy Maker, Legal Professional, AI Ethicist

Related on AIssential

• Most Dutch EU tobacco rule submissions were generated by a tobacco company’s AI tool — Corporate AI tools can subtly manipulate public consultations, generating biased responses at scale.
• Philip Morris rejects press claims it ‘rigged’ the EU tobacco consultation with AI — AI-assisted public consultation responses, particularly from corporate-backed platforms, raise concerns about democratic process integrity.
• Man used AI to make false statements to shut down London nightclub, police say — AI-generated false complaints are an emerging threat to regulatory processes and public administration.
• How AI and Robotics are Transforming the UK's Public Sector — AI and robotics are transforming UK public services and agriculture through data processing and biological solutions.
• Q&A: Do AI and bogus respondents threaten polling’s future? — The rise of AI-generated and faked responses in polling necessitates a return to rigorous, probability-based human sampling for trustworthy public opinion data.
• Stop Shouting. Start Policymaking. — AI-driven platforms can identify hidden public consensus, transforming divisive political discourse into actionable policy.

See Counsel's argued verdicts on the open AI decisions leaders are weighing →

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Welcome to the Artificial Intelligence Incident Database.