Campaigns Need AI Guardrails Before the 2026 Midterms

The political debate over artificial intelligence has focused on the easiest danger to understand: fake videos, fake voices and campaign ads that make voters question what is real.

That fear is justified with AI-generated political content already appearing in the 2026 midterm cycle, and experts have warned that deepfakes can blur reality, mislead voters and further weaken trust in elections. States have started to respond as The National Conference of State Legislatures has tracked a growing wave of state action on AI in elections and campaigns, especially around deepfakes and disclosure rules.

But if that is where the conversation stops, we are missing the larger problem.

Political AI will not confine itself to videos. It will write texts and fundraising scripts, answer voter questions, coordinate volunteers and shape the information campaign staff sees. The moment campaigns let AI handle direct voter engagement, the stakes change. A fabricated ad can be disowned. A flawed answer to a basic question, such as where to vote, whether a person is registered, or what identification is required, is far harder to detect and far more likely to push someone out of the democratic process.

Campaigns already use AI to draft emails, generate images or summarize data. That is the easy part. The harder frontier is when AI becomes the first point of contact for voters. If these systems are built carefully, they can help campaigns respond faster and listen better. If they are handled carelessly, they will overwhelm people with hollow language and confident errors that no human ever approved.

The real standard should be simple and far more serious: Campaigns must take responsibility for every message sent in their name. If AI drafts it, the campaign owns it. If AI answers a voter, the campaign owns that answer. If AI summarizes concerns, the campaign is accountable for how that information is interpreted and used. And when something goes wrong, the campaign must correct it.

Campaigns already sign off on their ads, mailers and fundraising appeals. AI does not change that burden. It only makes it easier to avoid it.

The Federal Election Commission declined in 2024 to open a new rulemaking specifically on AI in campaign ads, while noting that existing rules on fraudulent misrepresentation can still apply to deliberately deceptive AI-generated content. That leaves campaigns, parties and the technology companies that build these systems with a clear responsibility: Set the boundaries yourselves, because Washington is not going to do it for you any time soon.

These boundaries are not complicated. 

AI systems used in voter communication should work from campaign-approved materials and verified public information. They should not invent policy positions, give legal advice nor pull answers from the open internet. They should create records that campaign staff can review. And when voters ask sensitive or unexpected questions, the system should know when to stop and escalate to a person.

This is not resistance to innovation. It is the basic expectation for a technology entering an environment where public trust is already fragile. Americans doubt that institutions can regulate AI effectively, and that skepticism varies sharply by party. In that environment, campaigns cannot treat trust as something they can spend freely.

Calls to ban AI from campaigns entirely ignore political reality. The technology is here. Campaigns will use it because speed and scale decide elections. The question is whether they will use it as a responsible tool or as a shortcut that produces more noise.

Campaigns that mishandle AI will reveal themselves quickly. Their messages will feel generic. Their answers will avoid more than they engage. Their mistakes will appear careless. Voters may not know exactly why something feels off, but they will sense it.

Campaigns that use AI well will use it to understand people more clearly and move more efficiently, while keeping human judgment at the center.

The debate we should be having is not whether AI belongs in politics. The debate is whether campaigns are willing to be accountable for their use of it.

Tom Carroll is co-founder and CEO of Convos, a political technology company that helps campaigns manage voter conversations.