CiviClick Flooded Air Regulator With 20,000 AI-Generated Emails

The public comment process gives citizens a formal mechanism to shape regulatory decisions before they take effect. A Bloomberg feature investigation documents what happens when AI platforms industrialize that mechanism on behalf of paying clients: more than 20,000 emails opposing a proposed clean air rule arrived at Southern California's South Coast Air Quality Management District, generated by a Washington D.C.-based firm called CiviClick, which describes itself as "the first and best AI-powered grassroots advocacy platform." The district's board voted 7-5 to reject the proposal, which would have imposed fees on new gas-powered furnaces and water heaters across a region affecting roughly 10 million units in Orange County and parts of Los Angeles, Riverside, and San Bernardino counties.

The consent problem is the sharper issue. When the district's cybersecurity team contacted a sample of the listed commenters, three of the five who responded said they had no knowledge of any messages submitted in their name. A parallel campaign emerged at the Bay Area Air District, where an AI platform called Speak4 was used to submit comments targeting a separate air quality rulemaking. Advocates documented that at least 10 submissions were made without the supposed senders' consent, a finding the San Francisco Chronicle confirmed. One San Pablo resident described a letter filed under his name as "forged." Samuel Woolley, a researcher at the University of Pittsburgh, characterized the approach as "absolutely the next step in digital astroturfing," warning that AI "can make it look like people want things they actually do not want."

The reporting also cites CiviClick's use in a North Carolina gas pipeline campaign, with similar reports of messages that constituents said they never sent. What the investigation does not give you is a reliable count of how many regulatory proceedings are currently being targeted, or whether agencies have any systematic detection tools. Contacting a sample of commenters manually only surfaces the problem when someone happens to respond and say so.

New Hampshire has passed a disclosure requirement for AI use in political advertising, but state-by-state responses are trailing a practice that already spans multiple regulatory contexts. For regulatory affairs practitioners and organizations that depend on authentic grassroots input, the open question is how agencies should weigh a comment period when they cannot reliably verify which submissions are genuine.