AI is transforming the legislative process. An Arizona lobbyist talks about what he's seen so far

Artificial intelligence is becoming a bigger and bigger part of lots of aspects of our lives. And that’s no different at the Arizona Capitol.

If AI can help students write essays and home cooks come up with recipes, what’s to stop lawmakers from using it to write bills, for example?

Kevin DeMenna has been thinking about the ways in which AI might change how the legislative process works — both the potential pros and cons. DeMenna is a longtime lobbyist with DeMenna Public Affairs.

He talked to The Show about how much AI he’s observed being used in the legislative process over the last year or two.

Full conversation

KEVIN DEMENNA: Last year or two, it's become the issue with the most potential. Clearly it's not online in any significant way in the policy-making process, not officially for sure. The measure is that the speaker of the House just created a committee to be chaired by [Rep.] Justin Wilmeth, smart guy, the right guy to chair this committee that will take on AI issues, regulating it.

President [Donald] Trump's executive order apparently leaves states with very little authority, but the internal aspects, the impact at the Legislature, probably best to remind ourselves that our Constitution, it's slow, it's tedious, and the process is meant to be.

Now around that, what can you do? The bill production process, any of them can introduce really any number of bills. The production happens at a place called legislative counsel, and they are some of the smartest attorneys and drafters in the state.

And what they do is if you want to produce legislation, let's say for example, to tax pepperoni pizzas, you meet with them or you phone them, and in a matter of days out comes delivered to your office as a legislator, an introductory version of that bill.

But here's the fun part. We asked AI to draft a bill for us taxing pepperoni pizzas. And the inputs are critical. But after a couple of tries, we ended up with a 13-page bill that addressed some of the more nuanced aspects, like an effective date for the tax. Definition of pepperoni was kind of cool, but tucked in the middle of this bill was a reference to the Registrar of Contractors, which had no place, didn't belong there. So clearly humans have to preside over the end product.

MARK BRODIE: Well, that's kind of the key, right? That like yes, right now a bill that you ask AI to write is going to be imperfect. But the way that the technology is changing, it's not outside the realm of possibility to think that within some number of, maybe even months, like legislative counsel could be rendered obsolete, right. Like any legislator could just ask ChatGPT to write them a bill.

DEMENNA: And rules would have to be rewritten to accommodate that and the technology would have to be put in place. I know that they're considering it, but one of the things that's important to understand, particularly in our state government and at our Legislature: if it isn't broke, let's not fix it. And there are at least three study committees throughout state government that we found that are customer service, other aspects for AI.

In terms of the second area, these 90 legislators receive information from everywhere. As a legislator, your job isn't to become an expert on taxes or water or the content of a particular bill. Your job is to really find the smartest people, hopefully aligned with you politically, to advise you, to trust. But AI as it begins to feed information into the collective system, I don't have a forecast, except we have to be careful.

BRODIE: Well, I wonder, you talk about staff and the legislators themselves. I wonder if you think that AI might affect the kind of job that people like you do. Like if you are going to a legislator and trying to advocate for a bill or advocate against a bill or for an issue or against an issue, what's to stop a legislator from saying, "I don't need to talk to you, I'm just going to, I'm going to ask ChatGPT arguments for and against this, and then I'll make up my own mind."

DEMENNA: What an interesting scenario. I've been through a few generations of this before and after the Internet. And now with the arrival of AI, it's a terrifying experience to stand before a legislative committee with members that may be hostile to your issue. The best you can do is to hope to be the smartest guy on your issue, to be able to turn and hang at every corner.

The laptop in front of the member on the dais is now the smartest guy in the room. And there is an aspect to that where in real time, it's troubling because there aren't a lot of filters, but what it might just lead to is more in depth hearings, more detailed analysis because it's available.

BRODIE: So do you think then that maybe in the short term, the closer term, that the more applicable use for AI at the Legislature is sort of that information as opposed to drafting bills and coming up with actual policy? Is that a fair assessment, do you think?

DEMENNA: And I think that's the direction they're heading. So the answer is yes. It's a remarkable tool to put out the word on what they're doing, not just visually, but the deeper dives. My recommendation is if whatever app you have, try it if you're at all legislatively minded. I was amazed at the depth of the information, not amazed at the quality, but there was a lot.

BRODIE: So you referenced the new House committee dealing with AI, which seems to be an indication that some lawmakers believe this to be an issue that is worth sort of studying on its own. I wonder, sort of separate from how legislators use AI sort of in their day to day business, are there particular aspects of AI that you think the Legislature might be trying to regulate policy on or try to really delve into to understand better?

DEMENNA: We have legislative interest in cybersecurity, deep fakes. There is a track where what we've tried to regulate in social media. That perspective seems to be finding its way to AI issues. Candidly, the depth of knowledge at the Legislature on this issue has got to change.

BRODIE: Well, so how do you think this issue and this technology will continue to move forward in Arizona if, as you're suggesting, the folks who are in theory using it and creating policy for it maybe don't understand it as well as they ought to?

DEMENNA: I think we might have said the same thing about Twitter 10 years ago and other upgrades. I'm confident that anything that puts more information, and I recognize the issues of quality, veracity of what's being read, but more information is always better. Taking the time and being analytical in the approach, that's where this leads.

And if AI, maybe we should rename it, if trusted AI can accelerate that, then we're that much better off. We get lost going down rabbit holes. But if we could write an algorithm to do it for us. I'm curious what's down there.