AI Voter Texting Is Outpacing the Federal Law Meant to Govern It, CDT Says

AI texting moves into the mainstream of campaigning

CDT says 2024 may have been the first "AI election", but AI use in campaigns was still early and experimental then. Since that time, use of these tools has exploded, and an industry has grown up to supply them. According to CDT, they have been widely integrated into campaign technology in the 2026 midterms cycle.

One strategy is high-volume text messaging. Platforms such as aKillion, Convos and Vector Political are used by campaigns across the country to hold personalized one-on-one conversations with voters, adapting to responses in live time. CDT notes these exchanges can touch sensitive topics such as candidate preferences and voting history. Voters may not know they are texting an AI model, and may not know the conversation can be used to build a personalized voter file.

aKillion, which launched a platform called IONCandidate, says it turns incoming voter messages into a two-way, live-time conversation managed by AI. The platform replies in a candidate persona drawn from the candidate's voting record, official platform and additional approved messaging. aKillion claims sensitive and more complex conversations are "escalated to a human staffer." Convos uses an AI agent trained on candidate bios, platforms and policies to send starter texts. Campaigns are "encouraged" to disclose the AI agent.

What the TCPA covers, and what it does not

The TCPA, passed in 1991 and enforced by the Federal Communications Commission (FCC), was written to regulate prerecorded calls and mass robotexts. CDT says it was not designed for live, adaptive AI conversations that build an individual voter profile in real time.

According to CDT, the law requires the following:

• Prior consent for robotexts to a cell phone that use an autodialer.
• Prior express consent for outreach to a cell phone using a prerecorded or artificial voice, including AI-generated voice.
• Prior express consent for such voice outreach to a landline if contacted more than 3 times in a month.
• Identification of the calling or texting party whenever consent is required.

It does not require consent for political texts sent manually without an autodialer. It also does not require any disclosure that an AI or virtual assistant is being used, and it places no consent or disclosure obligation on campaigns or nonprofits using live human staff.

Unfinished and unclear FCC action

In 2024, the FCC issued a declaratory ruling that AI-generated voices are artificial, so their use in robocalls or robotexts requires the prior express consent of the called party. The guidance was prompted partly by a New Hampshire presidential primary incident in which voters received robocalls with a deepfake of President Biden's voice telling them not to vote.

CDT argues the ruling leaves text messages unclear, since texts often involve no "voice." An earlier Notice of Inquiry said the FCC's authority "encompasses current uses of AI in robocalling and robotexting," including interacting with consumers as though they were live human callers. According to CDT, the final declaratory ruling did not repeat the part on "interacting with consumers," and later FCC guidance on political calls and texts does not elaborate on text. CDT says this makes it hard for both the public and campaigns to know what is permitted.

The FCC also unanimously approved a Notice of Proposed Rulemaking in 2024 that would require callers and texters to disclose AI use and obtain prior written consent for AI-generated content. CDT says it has still not become a final rule more than two years later.

Separately, the FCC opened public comment on a petition from Club for Growth. It would remove the prior consent requirement for political robocalls to cellphones, including ones using AI voice, and give companies 10 business days to honor a do-not-call request. CDT says this raises serious concerns about bad actors using relaxed rules to deceptively target voters. According to CDT, the FCC could rule on the petition as soon as late October.

What it means for voters and campaigns

CDT's central point is that nothing in the TCPA, or in federal law more broadly, requires a campaign to say whether an AI agent or a human is on the other end of a text conversation. Vendors describe disclosure as something campaigns are "encouraged" to do, but CDT says encouragement is not a legal requirement, and voters have no guaranteed way to tell staff from a machine.

The piece also says it surveys an emerging patchwork of state disclosure laws and considers constitutional and privacy questions raised by regulating the technology. For campaigns, the ambiguity means uncertainty about their own responsibilities in text-based outreach.