An investigation by Pointer (KRO-NCRV) and NOS has uncovered large-scale public opinion manipulation in the Netherlands: approximately 75% of citizen responses to EU consultations regarding tobacco product regulation turned out to be products of generative AI. Of these AI responses, 95% were created using a specialized tool from Philip Morris International (PMI) as part of their 'Your Voice Your Choice' campaign.

image
image

What Happened

During EU consultations on tightening tobacco product regulations, it was discovered that the vast majority of citizen submissions were automated. A tool from Philip Morris International (PMI) allowed users to complete simple surveys, based on which an LLM generated personalized first-person appeals. These texts contained embedded arguments against the introduction of new restrictions, creating the illusion of mass support for the corporation's position.

Context

Technically, the process was implemented as a simple wrapper around a large language model. This case demonstrates the emergence of an AI-driven astroturfing industry, where generative AI is used not just for content creation, but for the strategic manipulation of institutional feedback mechanisms by imitating real citizen voices.

Why It Matters for the Industry

For the AI industry and developers, this sets a precedent for using technology to bypass democratic processes. It creates a need for the development of digital participation verification methods (Proof of Personhood) and tools to distinguish generated text from human text in mass feedback forms to protect government platforms from automated lobbying.

Why It Matters for Users

For the general public, this serves as a clear example of how AI tools can be used for covert and large-scale manipulation of public opinion under the guise of sincere civic initiatives, calling into question the legitimacy of such consultations.

What Is Not Yet Known / Limitations

The discussion is shifting from technical possibility to socio-political consequences and regulatory risks, but the exact scale of the influence on final EU decisions has yet to be assessed.

Sources

Author

Look at AI, Editorial Staff