AI Bots Tried to Sabotage Disposable Vape Policy, Analysis Finds
Artificial intelligence is often discussed in terms of dramatic risks: bioweapons, cyberattacks, or systems that behave in ways their creators did not intend. But a quieter form of AI misuse has now surfaced in the middle of a public health debate. According to a new analysis published in a medical journal, automated bots were used to flood a formal consultation on disposable vapes, attempting to create the impression that ordinary people overwhelmingly opposed restrictions. The episode is being described as the first documented case of large-scale automated interference in a public health consultation.
Key Facts
- A 2023 public health consultation on disposable vapes and related policies received 118,756 responses.
- About 76% of those responses were eventually flagged as fraudulent or bot-generated.
- Of the suspicious submissions, 90,833 out of 90,835 answered only one question: whether to restrict disposable vape sales.
- The bot responses overwhelmingly opposed restrictions, raising suspicion among researchers.
- After the fraudulent answers were removed, 79.6% of the remaining respondents supported restrictions on disposable vape sales.
- Researchers call the episode the first documented large-scale automated interference in a formal public health consultation.
- The pattern resembles coordinated tobacco industry lobbying, though the origin of the bot campaign has not been identified.
A Consultation Flooded by Automated Responses
The consultation asked the public for views on disposable vapes and other health topics. Participants submitted responses through multiple-choice questions, a format designed to gather broad public input before policy decisions. In theory, such consultations help regulators understand public sentiment, identify concerns, and build legitimacy for new rules. In practice, they depend on the assumption that the people responding are real people with genuine views.
That assumption appears to have broken down. Researchers found that the vast majority of submissions were not ordinary public comments at all. The volume alone was notable, but the distribution of answers was more revealing. Tens of thousands of responses addressed only the question about restricting disposable vape sales and skipped every other issue. That behavior suggested a targeted campaign rather than organic participation. When researchers analyzed the submissions, they concluded that most were automated or fraudulent.
The finding matters beyond the specific question of vapes. Public consultations are a core part of how modern governments test policy, especially in health. If automated accounts can distort those processes, they can weaken the evidence base for regulation, create false impressions of public opposition, and make it harder for officials to distinguish genuine concern from manufactured noise.
Why Disposable Vapes Are a Public Health Flashpoint
Disposable vapes have become a major focus for public health officials because of concerns about youth nicotine use, addiction, and environmental waste. The devices are cheap, easy to conceal, and often sold in bright flavors that critics say appeal to young people. Many health advocates argue that stricter rules are needed to prevent a new generation from becoming dependent on nicotine. Others warn that overly broad restrictions could push adult smokers back toward cigarettes or create unregulated markets.
In the U.K., the debate has been particularly intense. The government has considered restrictions on disposable vapes as part of a broader effort to reduce youth vaping and create a smoke-free generation. The consultation in question was run by the health department and drew a massive public response. That level of engagement is normally a sign of democratic health. In this case, however, the numbers were inflated by what researchers describe as coordinated automated activity.
The policy stakes are high because vaping regulation sits at the intersection of addiction, public health, commercial interests, and personal freedom. A consultation that appears to show strong opposition can influence how ministers frame the issue, how quickly they act, and how confident they feel about enforcing new rules. When bot activity distorts that picture, it can affect the direction of policy even if the underlying public health evidence remains unchanged.
The Pattern Looked Coordinated, Not Organic
Researchers did not simply notice a large number of responses. They noticed a pattern. The submissions were overwhelmingly opposed to restrictions, and they were concentrated almost entirely on the disposable vape question. That directional consistency is important. Real public consultations usually produce a mix of views, varying levels of detail, and uneven participation across questions. A flood of identical or near-identical answers, especially on a single high-profile issue, is a warning sign.
The concentration was extreme. According to the analysis, 90,833 out of 90,835 responses answered just one question. That means the campaign was not a broad attempt to engage with the consultation. It was a narrowly aimed effort to influence one specific policy question. The scale and consistency led researchers to conclude that the activity was coordinated rather than spontaneous.
This does not necessarily mean the people behind the campaign used sophisticated AI. Simple automation can generate large volumes of text and submit responses at scale. But modern AI tools make such campaigns cheaper, faster, and more adaptable. A bot can produce varied phrasing, avoid simple duplication filters, and mimic the tone of a concerned citizen. That makes detection harder and raises the risk that future campaigns will blend into genuine public debate.
An Old Lobbying Playbook With New Tools
The episode echoes earlier efforts by commercial interests to shape public consultations. Tobacco companies and other industries have long been accused of encouraging mass coordinated submissions, sometimes through campaigns that present industry talking points as grassroots concern. The tactic is often called astroturfing: creating the appearance of spontaneous public support or opposition when the activity is actually organized and funded.
What is new is the use of automation and AI to scale up that playbook. In the past, a lobbying campaign needed people to write letters, sign forms, or submit comments. That took time and resources. Automated systems can now produce thousands of responses in a fraction of the time, with enough variation to avoid obvious detection. The result is a form of synthetic public opinion that can be used to pressure policymakers, confuse journalists, and undermine trust in legitimate consultation results.
The researchers noted that the scale and directional consistency of the bot responses suggested a coordinated campaign. They also pointed to prior evidence of tobacco industry efforts to manipulate public consultations through mass coordinated submissions. The comparison is uncomfortable because it suggests that the same basic strategy has simply been upgraded for the AI era. The tools are new, but the goal is familiar: make it look as if the public opposes regulation.
Why Consultations Are Vulnerable to Bot Interference
Public consultations are designed to be open. That openness is a strength, but it is also a weakness. Many systems allow anyone to submit a response without verified identity, and some allow multiple submissions from the same person or device. That makes them easy to target. A campaign does not need to hack a government database. It only needs to fill in a web form repeatedly.
Defenders of open consultations argue that barriers to participation can exclude marginalized voices and reduce the quality of public input. That concern is real. But the vape case shows that openness without safeguards can also produce distorted results. If a single coordinated campaign can generate three-quarters of all responses, the consultation stops being a reliable measure of public opinion.
Possible safeguards include stronger identity verification, limits on submissions per device or account, anomaly detection, and statistical checks for suspicious patterns. Governments could also publish more detailed methodology about how responses are screened and weighted. None of these measures is perfect, but together they could make it harder for automated campaigns to drown out genuine participants.
The Wider Risk for AI and Public Health
The vape consultation is a reminder that AI risks are not limited to science-fiction scenarios. Much of the public debate focuses on extreme dangers such as bioweapons or critical infrastructure attacks. Those risks deserve attention. But AI can also be used in subtler ways to manipulate democratic processes, influence health policy, and erode trust in public institutions.
Health policy is especially vulnerable because it often involves emotionally charged issues, commercial interests, and scientific uncertainty. A well-timed bot campaign can amplify doubts, create the impression of public backlash, and give policymakers an excuse to delay action. Even when the campaign is eventually exposed, the confusion it creates can linger. Once false claims enter the public record, they can be repeated by media outlets, social media users, and industry groups long after the original submissions are removed.
The researchers were unable to pinpoint who was behind the automated responses. That is another challenge. Attribution is difficult in online campaigns, especially when bots are deployed across borders or through commercial services. Even without attribution, the effect on public trust can be significant. Citizens may begin to doubt whether their own participation matters if consultations can be gamed so easily.
What the Numbers Show After Bot Removal
The most striking finding emerged after the fraudulent responses were discarded. Once the bot submissions were removed, a strong majority of remaining respondents supported restrictions on disposable vape sales. The figure was 79.6%. In other words, the apparent public opposition was not just exaggerated. It was reversed. The manufactured response created a false impression that most participants opposed restrictions, when the genuine responses pointed in the opposite direction.
That reversal is important for policymakers. It suggests that the consultation, if analyzed carefully, did not show public resistance to stronger vape rules. It showed the opposite. The episode also demonstrates why raw response counts should never be treated as a direct measure of public opinion. Volume is not the same as legitimacy, and a large number of submissions can be a sign of manipulation rather than engagement.
Safeguards Needed for Digital Participation
The case is likely to accelerate calls for better protections around digital public participation. Governments already use online forms for consultations on housing, transportation, taxation, and health. As those processes become more digital, they will become more attractive targets for automated manipulation. The same tools that make it easy for citizens to participate also make it easy for bots to participate at scale.
One approach is to treat consultation data the way election officials treat voting data: with layers of verification, auditing, and transparency. That could include publishing aggregate results alongside information about suspicious activity, using forensic analysis to identify coordinated campaigns, and giving researchers access to anonymized data so independent experts can check the findings. Another approach is to combine open consultations with representative surveys, focus groups, and expert evidence, so that policy is not driven by raw online volume alone.
AI developers also face pressure to build stronger safeguards against misuse. Text-generation systems can be used to produce persuasive messages at scale, and detection tools often lag behind the latest models. Some companies have added watermarking, usage monitoring, and restrictions on harmful applications. But the vape consultation shows that even relatively simple automation can have an outsized effect when it targets a narrow policy question with a high-stakes outcome.
The analysis does not identify who orchestrated the submissions. It does show that automated text can now reach the heart of the policy process, where a single question can shape how millions of people live, breathe, and manage addiction. For public health officials, the lesson is clear: the next consultation may look like a broad democratic exercise while being quietly engineered by machines.
Source: Digital Trends News