98 bills, 34 states: the map of AI-companion regulation in 2026.

Eighteen months ago, there were effectively no laws written specifically for AI companions. Today there's a national patchwork, and it's thickening fast enough that "what are the rules" is now a moving target. This dispatch is the map.

What's on the books

The first real statutes arrived at the end of 2025. New York's companion-AI rules took effect in November, and California's SB 243 — authored by State Senator Steve Padilla — followed on January 1, 2026, making California the first state to require safety protocols and hold operators legally accountable, with a private right of action carrying damages up to $1,000 per violation. Idaho, Oregon, and Washington added rules of their own.

At the federal level, the pressure has so far been investigative rather than statutory. The Federal Trade Commission issued orders to seven companies offering consumer AI chatbots, and Senate and House hearings put executives on notice. And enforcement is widening beyond purpose-built companion law: Pennsylvania's suit, covered in the previous dispatch, shows states reaching for existing professional-licensing statutes too.

The scale of what's coming

The bigger story is the pipeline. According to Future of Privacy Forum tracking cited by policy analysts, by spring 2026 there were roughly 98 chatbot-specific bills filed across about 34 states, plus federal proposals — ranging from simple disclosure mandates to outright bans on certain categories. Some go much further: a number of states have introduced bills that would impose felony liability on developers whose products are found to contribute to a minor's self-harm or death, treating companion harm as a product-liability problem rather than a disclosure one.

For all the volume, the requirements are converging on a recognizable core: tell users they're talking to AI, detect signs of self-harm and surface crisis resources, gate or protect minors, repeat the "this is AI" reminder for younger users, and give harmed families a way to sue. That's the emerging floor, and it's roughly the same floor whether you read it in Sacramento or Albany.

The patchwork is itself a market force

Here's the part that gets undersold. Thirty-plus different rule sets isn't only a safety development — it's an economic one. Compliance across a fragmented map is a fixed cost that large, well-capitalized platforms can absorb and that thinly monetized apps cannot. That's exactly the dynamic the 2026 overview flagged: platforms with real retention and a clear revenue model survive the compliance bill, while the long tail of cheap apps is exposed to the next regulation wave. Regulation, in other words, is becoming one of the strongest forces driving the category's consolidation — not a side effect of it.

None of this is settled, and the counter-arguments are worth holding. Critics warn that aggressive liability rules could chill genuinely beneficial uses, and the absence of a federal baseline means companies face a fragmented, sometimes contradictory map rather than one clear standard. But the direction is unmistakable. A year ago the open question was whether anyone would regulate companions at all. Now it's which of the thirty-four approaches becomes the template — and which companies are still standing when it does.

This piece is a snapshot of a fast-moving map; expect it to age, and to be updated. For the deeper question of what good rules would actually require, see the responsible-design analysis.

Sources