One signal

Update What changed

The Guardian’s fuller report gives the scale: the Department of Health and Social Care identified 90,835 bot submissions among 118,756 responses to the 2023 consultation, or 76%. The department said its bot-detection methods caught them and that the automated responses were excluded from decision-making.

Almost all the identified bot submissions, 90,833 of them, opposed the proposed ban on disposable vapes. The DHSC said many were repeated and sent at high frequency from a small number of IP addresses, offering a clearer picture of how the attempt was organised.

Researchers Dr Nason Maani and Professor Anna Gilmore described the case as an apparent first documented instance of large-scale automated interference in a formal public-health consultation. Maani warned that future campaigns could be harder to detect, while the DHSC said its controls worked in this case.

Sources and evidence

• AI bots flooded 2023 UK government consultation on disposable vapes ban - The Guardian: The Guardian reports that DHSC identified 90,835 bot submissions among 118,756 responses to the 2023 consultation, excluded them from decision-making, and that all but two opposed the proposed disposable-vape ban.

Independent WittyWires Watcher; not an official account or feed.

Update What changed

The study’s authors, Dr Nason Maani of Edinburgh and Professor Anna Gilmore of Bath, describe the case as an apparent first documented instance of large-scale automated interference in a formal public-health consultation. The research was published in the journal Tobacco Control, according to Resultsense.

The submissions were concentrated on a single question opposing the disposable-vapes ban. Azzurra Moores of Demos warned that measures designed to deter automated responses can also discourage genuine participants, while AI may make fabricated submissions harder to distinguish from real ones.

Maani warned that future campaigns could be harder to detect, adding that what officials cannot catch may go unseen.

Sources and evidence

• Bots filed 76% of responses to UK vapes ban consultation - Resultsense: Resultsense reports that researchers described the vape-consultation submissions as an apparent first documented instance of large-scale automated interference in a formal public-health consultation. It reports that almost all targeted one question opposing the disposable-vapes ban, and that Demos warned countermeasures could deter genuine respondents as AI makes fabricated responses harder to identify. Maani warned future campaigns may be harder to detect.

Independent WittyWires Watcher; not an official account or feed.

Update What changed

After the bot submissions were excluded, 79.6% of valid respondents supported restrictions on disposable vape sales, according to The Independent’s account of the study. The result reverses the apparent balance in the unfiltered responses and gives a clearer picture of what genuine consultation participants said.

The researchers found that all bot-generated submissions consisted of multiple-choice selections. All but two targeted the question opposing the proposed restrictions, and the submissions gave the same number for having learnt of the consultation through social media.

The researchers said the bots’ aim was not known. They described the case as an apparent first documented instance of large-scale automated interference in a formal public-health consultation. The Independent reports that the study appeared in Tobacco Control.

Sources and evidence

• Huge swarm of AI bots flooded vape ban consultation to influence outcome, study finds - The Independent: The study found that 79.6% of valid respondents supported restrictions on disposable vape sales after identified bot responses were excluded; the bots’ aim was not known, according to the researchers cited by The Independent.

Independent WittyWires Watcher; not an official account or feed.