State Affairs raises $70M for AI policy intelligence platform

Every few years a company tries to build the Bloomberg of something-not-finance, and most of the attempts quietly fade. State Affairs is the newest attempt at policy, and what makes the announcement in its PR Newswire release worth pausing on is not the $70 million or the investor list, it is the shape of the moat they are describing.

The pitch, in plain terms, is that most AI systems in this space are limited by what is already published, and state capitols are one of the least covered, most fragmented data sources in American life. The company's own number is that state legislatures introduced more than 135,500 bills in 2025, up roughly 55% from 87,500 the year before. Their answer is to run an actual newsroom, more than 2,000 originally reported nonpartisan articles a month according to the release, and feed that reporting plus structured government data into an AI layer that tracks every bill, regulation and hearing across all 50 states and the federal government.

Why this matters if you are not in government relations: policy and regulatory exposure is starting to be priced like a real enterprise data category, next to market data and threat intel. The customer list they are willing to name, Walmart, Mastercard and McDonald's, is the tell, along with the claim that the platform is in active use by roughly one-third of state and federal elected officials. Co-founder and CEO Evan Burns frames it as, and I am quoting from the release, "Policy and regulatory markets are often more impactful to organizations than" financial markets. The investor mix, Founders Fund, Khosla Ventures, Tru Arrow Partners, Alumni Ventures, plus Marcus Brauchli, The Athletic's Alex Mather and Adam Hansmann, and KKR's Richard Sarnoff, is unusually media-literate for a venture round.

The honest caveat is that a press release is a press release. It does not disclose the actual reporter headcount, the LLM stack under the hood, how "nonpartisan" is enforced as the newsroom grows, or how the one-third-of-officials figure is measured. An AI layer summarizing that many bills will get things wrong sometimes, and a legislative staffer acting on a bad summary is a real failure mode. Take the specifics as reported, not settled.

The part worth watching is whether original reporting turns out to be a defensible moat for AI products. If it does, this is a template a lot of underserved beats, courts, procurement, municipal budgets, could copy. If it does not, the same enterprises will end up paying whoever scrapes the fastest.