Generated by Codex with GPT-5

This summary covers The Economist’s April 25th, 2026 United States article listed in the contents as Legislating with AI and published under the headline Vote for Claude.

The article argues that artificial intelligence is quietly becoming part of the machinery of American state government. The change is not arriving first as grand constitutional theory or futuristic campaign rhetoric. It is arriving through overworked legislators, small staffs and mundane bottlenecks: research, bill drafting, talking points, hearing preparation and quick checks against lobbyists’ claims.

That makes the story both practical and unsettling. AI can help weakly resourced lawmakers do their jobs better. It can also tempt them to outsource parts of judgment that voters expected them to exercise themselves.

Why Statehouses Are Turning To AI

The article opens with Kent Roe, a Republican state representative in South Dakota who is also a farmland appraiser, utility-board member and church-council participant. South Dakota’s legislature meets for only part of the year and has a tiny professional staff by national standards. For legislators like Roe, AI is not a novelty. It is a way to stretch scarce time.

Roe uses chatbots for research, argument-building and early legislative drafting. He runs ideas through Grok to see whether other states have passed similar laws and to test whether a proposal may have constitutional weaknesses. His view is utilitarian: AI is a tool, much like a calculator or a phone. The article uses him to show why adoption is spreading fastest in places where the need is obvious. Many state lawmakers are part-time officials expected to master complicated subjects with far less support than members of Congress or executive agencies receive.

The evidence points to rapid uptake. A survey by the National Conference of State Legislatures found that 44% of state legislative staff used AI in their work last year, up from 20% in 2024. Governors and state agencies are also experimenting with the technology, including as a way to simplify or prune poorly written regulations.

The appeal is easy to understand. AI can compress hours of background research into minutes. It can summarize policy debates, surface examples from other states and help a legislator prepare questions. For lawmakers facing better-funded lobbyists, that matters. The article quotes a Vermont Democrat who says she uses AI during committee hearings to fact-check lobbyists in real time, and has caught them making false claims.

The Slop Risk

The article does not treat AI as magic expertise. Its main technical warning is familiar but important: general-purpose models invent things. That is especially dangerous in lawmaking, where a fake statute, misquoted precedent or nonexistent case can distort a debate or waste staff time.

Roe tries to manage that problem by running prompts repeatedly to see whether the answer changes. Another lawmaker says hallucinations are especially common when models discuss case law. The formal process still has safeguards: legislative lawyers, editors and researchers review bills before they move forward. But those safeguards are already under strain. If AI makes it easier to generate more draft bills, amendments and talking points, human reviewers may face a larger volume of low-quality material.

That is the “slop laws” concern. The danger is not simply that a chatbot writes a bad sentence. It is that legislatures with limited staff may become flooded with plausible-looking drafts that require careful checking. AI lowers the cost of producing legislative text, but it does not lower the cost of making law responsibly by the same amount.

There is also an asymmetry in how errors travel. A legislator may use AI to move faster, but an incorrect citation or misleading comparison can become part of a political argument before anyone has time to verify it. In a statehouse, as in a courtroom or newsroom, speed is valuable only if the output can still be trusted.

The Democratic Question

The deeper issue is not accuracy alone. The article asks what happens to representative judgment when lawmakers rely on chatbots for tasks that are part of thinking itself. Writing, problem-solving and argument-building are not clerical extras. They are how legislators clarify what they believe, identify trade-offs and decide what a bill should actually do.

Used well, AI can support that work. It can supply examples, reveal overlooked objections and help a legislator test a proposal before committing to it. Used badly, it can become a substitute for independent thought. A lawmaker who asks a model to generate the argument, refine the bill and anticipate criticism may become less the author of a policy than the operator of a prompt.

The article’s strongest point is that this matters because voters do not elect software. They elect people who are supposed to be accountable for choices. If AI helps a legislator understand a subject more quickly, democracy may benefit. If it lets legislators sound prepared without doing the hard work of judgment, democracy may become thinner.

That distinction will be difficult to police. Legislatures can require disclosure, issue guidelines or restrict certain uses, but the most important line is behavioral: whether officials treat AI as an assistant or as an authority.

The Takeaway

The Economist presents AI in state legislatures as an early test of how ordinary democratic institutions absorb powerful tools. The technology is attractive because state government is often understaffed, slow and vulnerable to lobbyists with deeper expertise. In that setting, AI can make public officials more capable.

But lawmaking is not just information processing. It is judgment under pressure, with consequences for citizens who cannot inspect every prompt or verify every draft. The useful future is one in which AI gives legislators more capacity while leaving responsibility clearly with the humans voters chose.

In plain English: AI may help lawmakers work faster and push back against better-resourced interests. It should not become a way for them to stop thinking for themselves.