> ## Content Index
> Fetch the complete content index at: https://aiaffairs.nz/llms.txt
> Use this file to discover other available public pages before exploring further.

# Maybe politicians should be required to have an AI we can interrogate
- URL: https://aiaffairs.nz/maybe-politicians-should-be-required-to-have-an-ai-we-can-interrogate/
- Published: 2026-08-29T00:00:00.000Z
- Updated: 2026-09-20T23:41:44.000Z
- Author: Mehrdad Fatemi
- Tags: Editorials, Policy & Regulation, From the Publisher

A 2026 Collective Intelligence Project survey found 55.9% trusted their AI chatbot to act in their interests, versus 32.6% for elected representatives

The Bloomberg piece on AI-generated political influencers in Brazil got me thinking that we may be heading towards the wrong regulatory response.

The obvious reaction to AI in politics is to restrict it.

Don’t let it recommend candidates. Don’t let it impersonate people. Don’t let it answer certain questions. Don’t let it influence voters.

Some of that is obviously necessary. Deepfakes and fabricated endorsements are fraud with better tooling.

But I think banning political AI from answering people’s questions is the wrong instinct.

I would almost go the other way.

Every politician should have an official AI.

Give it a name. Register it. Make it clear who owns it. Make the politician legally responsible for what it says.

Then let people interrogate the hell out of it.

Ask it what the policy actually costs.

Ask who pays.

Ask who loses.

Ask what happens if the assumptions are wrong.

Ask it why the politician said the opposite five years ago.

Ask the same question twenty different ways until the spin stops working.

That would be far more useful to democracy than another carefully managed press conference.

Politicians are remarkably good at not answering questions. That is almost a professional competency.

They have media training, talking points, time limits, press secretaries and the ability to answer a different question from the one they were asked.

An AI does not have to be allowed that luxury.

If it is the politician’s official AI, it should be grounded in their manifesto, speeches, voting record, policy documents and public commitments.

And its answers should be recorded.

That is the important bit.

I’m not actually convinced AI should become a legal person. That sounds clever until the first politician says “my AI made a mistake”.

Congratulations, we’ve just invented a new liability shield.

I’d rather treat it as an agent.

If your AI says something, you said it.

If it gets something wrong, you correct it publicly.

If its position changes, the change is visible.

If two versions of your AI give two different answers to two different groups of voters, we should be able to see that too.

In fact, that may be one of the most useful things AI could do to politics: make inconsistency much harder.

The public fascination with “AI politicians” is also telling.

I hear people in different countries say they would rather have an AI than some of the incompetent, corrupt or purely performative politicians they currently have.

Usually half joking.

But only half.

I don’t think most people actually want an algorithm running the country.

I think they want something they can question.

Something that remembers.

Something that can explain why a decision was made.

Something that cannot pretend it never said something.

Something that does not need six advisers to prepare an answer to a fairly basic question.

That is quite different from handing political power to AI.

So perhaps the real regulatory question is not whether an AI political personality is “real”.

Who cares?

The useful question is:

Who is accountable for what it says?

If the answer is a real politician, party or campaign, with real legal liability, then I am not particularly frightened by the idea of political AI.

I am much more interested in what happens when every politician has a permanently available representative that the public can interrogate without permission.

That could be uncomfortable.

Which is probably why it might be useful.

  
---

*Originally published on* [*LinkedIn*](https://www.linkedin.com/pulse/maybe-politicians-should-required-have-ai-we-can-mehrdad-fatemi-prcne/?ref=aiaffairs.nz)*.*