AI should build trust, not replace it
People will push back on political AI. Groups that put ethics first, and can prove it, will keep public trust.
Every generation of campaign technology arrives the same way. First it is a curiosity. Then it is an advantage. Then it is a scandal. Direct mail, television advertising, micro-targeted social advertising — each followed the arc, and each ended with a regulatory settlement written by people who were angry about how the technology had been used by its least careful practitioners.
Generative AI is moving through that arc faster than anything before it. Deepfakes, disinformation and synthetic media are already reshaping public trust, and the public conversation about political AI is being shaped right now — largely by its worst examples.
The asymmetry nobody is pricing in
Here is what most organisations deploying AI in public communication have not worked out. The regulatory settlement is not the risk. The risk is the day a journalist asks you, on camera, to account for something you generated eighteen months ago — and you cannot.
Technology without accountability is indistinguishable from manipulation. Not rhetorically indistinguishable: operationally indistinguishable. If you cannot produce the provenance of a piece of media, you cannot prove you are not the thing you are being accused of being.
The question isn't whether AI will transform politics. The question is whether we'll lead that transformation with integrity, or follow it with regret.
What leading actually requires
It requires four things, and all four have to be in place before deployment rather than after the first incident.
- Content watermarking. Every AI-generated image, video or voice clip embedded with C2PA content credentials showing provenance, creation date and synthetic nature.
- A public transparency portal cataloguing all AI deployments, usage statistics and quarterly audit reports — searchable and downloadable by anyone.
- Data handling that exceeds minimum compliance: explicit consent, easy opt-out, regular deletion, and citizen access to their own records.
- Independent oversight with the power to review any deployment and recommend changes, reporting publicly and unredacted.
The fourth is the one organisations resist, and it is the one that matters. An ethics council that cannot say no is a communications asset, not an oversight body — and the distinction becomes obvious to the public at precisely the wrong moment.
The strategic advantage
There is a cynical case for all of this, and it is worth stating plainly because it is also true. When the inevitable backlash against political AI arrives, the organisations that can credibly demonstrate they led with ethics will be the only ones left with a licence to operate.
But the better case is simpler. These are tools for connecting institutions to the people they represent. Used well, they let a government hear a citizen in a village it has never polled, and answer them in the language they speak at home. That is not a threat to democratic tradition. It is a restoration of something broadcast communication took away.
Innovative but principled. Forward-looking but rooted in democratic values. Efficient but never cynical. That is not a compliance posture. It is a description of what the technology is actually for.
