Purported Philip Morris-Backed AI Tool Reportedly Generated Responses Opposing Stricter EU Tobacco Rules

Philip Morris International reportedly promoted a "Your Voice Your Choice" tool that allegedly generated first-person responses for an EU call for evidence on tobacco regulation. Investigators reportedly linked hundreds of submissions to the tool and found that it allegedly largely steered users toward arguments against stricter rules, sometimes by purportedly adding points they had not selected.

Risk classification

• Primary risk domain: 4 Malicious actors
• Primary risk subdomain: 4.1 Disinformation, surveillance, and influence at scale

The incident involves a coordinated, large-scale campaign by a corporate actor using AI to manipulate a public consultation and influence EU political processes.

Additional risk subdomains

• 3.1 False or misleading information: The AI tool generated fabricated citizen responses and, in some cases, inserted arguments that users did not actually select.

Causal factors

• Entity: Human
• Intent: Intentional
• Timing: Post-deployment

The incident was caused by the intentional decision of humans at Philip Morris International to deploy an AI tool to manipulate a public consultation.

EU AI Act risk tier

• Risk tier: 3 Limited Risk

Limited Risk: The system is an AI-enabled text generator used to produce content, which triggers transparency obligations under the Act to ensure users and recipients are aware the content is AI-generated.

AI system and alleged parties

• AI system: Your Voice Your Choice
• AI purpose: Writing Assistant; Technical Text Generation
• Behaviour type: Assistant
• Alleged developer: Tobacco companies, Philip Morris International, Large language model developers, AI-assisted public consultation response generator developers
• Alleged deployer: Tobacco companies, Philip Morris International, Information manipulation actors
• Alleged harmed parties: Public health, Public consultation integrity, Participants in European Commission public consultations, General public of the European Union, General public, European Commission, Epistemic integrity, Democratic integrity

Harm severity

Severity ratings are based on this scale.

Highest direct severity in any category: Minor. Severity is scored from Negligible to Catastrophic in each harm category, for harm the reports describe as caused directly or indirectly by the AI system.

• Physical: direct Negligible, indirect Negligible
• Infrastructure: direct Negligible, indirect Negligible
• Property: direct Negligible, indirect Negligible
• Financial: direct Negligible, indirect Negligible
• Environmental: direct Negligible, indirect Negligible
• Malicious content: direct Negligible, indirect Negligible
• Differential treatment: direct Negligible, indirect Negligible
• Civil rights: direct Negligible, indirect Negligible
• Democracy: direct Minor, indirect Minor
• Privacy: direct Negligible, indirect Negligible
• Psychological: direct Negligible, indirect Negligible
• Epistemic: direct Minor, indirect Minor
• Child sexual exploitation and abuse: direct Negligible, indirect Negligible

Democracy

Reported: The report explicitly describes an attack on democratic integrity and manipulation of an official public consultation.

Directly caused: The AI tool generated 22,500 pre-formatted responses (35% of analyzed contributions) to influence an EU public consultation on tobacco regulation, undermining the integrity of citizen participation.

Indirectly caused: The campaign attempted to delay or weaken public health legislation by saturating the consultation with automated responses, potentially affecting EU-wide tobacco policies.

Inferred additional harm: The erosion of trust in public consultations and democratic processes within the EU, potentially leading to stricter verification measures or disregard for genuine citizen feedback.

Epistemic

Reported: The report explicitly describes epistemic harm through the fabrication of responses and misrepresentation of citizen opinions.

Directly caused: The tool generated first-person responses representing commercial interests as genuine citizen feedback, and in at least 50 cases, added arguments against regulation that the users had not selected.

Indirectly caused: The pollution of the public consultation database with 22,500 AI-generated responses, distorting the perceived public consensus on tobacco regulation.

Inferred additional harm: Increased difficulty for policymakers to distinguish between genuine human opinions and automated lobbying campaigns, degrading the information ecosystem of policy-making.

People affected

• Occurrences reported: 1
• People reportedly harmed: 50
• People reportedly exposed: 22500

Potential causes

Management

• Prioritizing Profit Over Integrity: PMI management deployed AI to protect revenue from heated tobacco and vapes.
• Concealed Corporate Lobbying: Management hid corporate identity behind the 'Your Voice Your Choice' campaign.

Technology

• Biased Argument Generation: The AI tool automatically added anti-regulation arguments not chosen by users.
• Automated Text Generation: AI generated unique first-person text to bypass duplicate submission filters.

Data Inputs

• Manipulative Questionnaire Options: Eleven options presented to users mostly supported anti-regulation arguments.
• User Input Distortion: The tool used user profile data to synthesize biased personal statements.

Human Factors

• Deceptive Branding Misleading Users: Campaign hid PMI's involvement, leading users to unwittingly support lobbying.
• Mobilization of Consumer Networks: Industry mobilized consumers to submit AI responses without full awareness.

Process and Methods

• No AI Detection on EU Portal: The EU Commission lacked mechanisms to filter or flag AI-generated responses.
• Remote-Controlled Response Campaign: Pre-formatted automated structures left little room for genuine personal views.

Regulatory Environment

• Lack of Public Consultation Rules: No guidelines existed to prevent AI-driven lobbying in EU public consultations.
• Circumvention of WHO FCTC Rules: PMI bypassed Article 5.3 rules prohibiting industry influence on health policy.

Information quality

• Classification confidence: High
• Reason for confidence: The reports provide precise statistics, such as the 22,500 AI-generated responses and the 68% figure for France, backed by a collaborative investigation by major Dutch public broadcasters.
• Ambiguities identified: The exact technical architecture of the AI tool is not specified beyond being a generative AI tool.
• Alternative interpretations: None, the evidence of a coordinated corporate lobbying campaign using AI is highly clear and uncontested.

Philip Morris International deployed an AI tool to generate over 22,500 fake citizen responses to an EU tobacco regulation consultation. This represents a highly coordinated AI-driven astroturfing campaign designed to distort democratic policy-making, demonstrating how generative AI can scale corporate influence operations to undermine regulatory processes.

• Overall national security impact: Minor
• Scope: Multiple nations
• Primary target: European Union
• Other affected: France, Netherlands
• Alleged perpetrator: Philip Morris International

Threat characteristics

• Imminence: Long-term. Represents a strategic capability concern regarding AI-enabled astroturfing rather than an active crisis requiring immediate intervention.
• Autonomy: Human-supervised. The AI generated text based on human inputs but acted with some autonomy by injecting unselected arguments without user consent.
• Novelty: Evolved capability. Represents a significant advancement of existing corporate lobbying and astroturfing techniques through the scale of generative AI.

Impact by dimension

• Physical security: Negligible. The incident did not involve any threats to physical systems, critical infrastructure, or human safety.
• Information security: Minor. Coordinated deployment of generative AI to fabricate 22,500 citizen responses, representing a sophisticated corporate influence campaign targeting EU policy-making.
• Sovereignty: Minor. The campaign targeted an official EU public consultation, attempting to distort democratic feedback and policy-making through automated, non-consensual submissions.
• Economic security: Negligible. No threats to financial systems, strategic technology theft, or critical supply chain disruptions were reported.
• Societal stability: Minor. The manipulation of citizen feedback risks eroding public trust in democratic institutions and consultation mechanisms, though it did not cause civil unrest.