14 States, One Deadline: AI Companion Chatbot Laws, Verified

Why I rebuilt this count myself

I sat down to write this post with what I thought was a simple job: cite the going count of states with companion-chatbot laws on the books, point at the nearest deadline, and move on. Instead I spent two days re-checking every bill against its own primary source, because the two trackers I follow most closely — MultiState and the Transparency Coalition — don't agree with each other, and neither of them agreed with what I could find by just reading the bill text and the governor's signing record myself. This is the same discipline Matterwire's own regulatory change management product applies to every record it surfaces — a source URL and a capture timestamp, not a summary of someone else's summary.

Here's where I landed: as of today, 14 states have signed companion-chatbot safety measures into law in 2026, and six of them share the exact same effective date — January 1, 2027. That's the number I'm putting my name on. Below is exactly how I got there, so you can check my work instead of taking my word for it.

Why the trackers disagree

MultiState's own tracker puts the number around 12 states. The Transparency Coalition's 2026 Mid-Year report counts 14 measures across 13 states. I initially assumed this was purely a definitional gap — and it partly is — but once I checked every bill individually, I found two more reasons the numbers move:

1. Definition. MultiState counts only bills whose primary subject is companion or AI-interaction chatbots. The Transparency Coalition also counts a chatbot-specific provision sitting inside a broader AI bill — which is how Wyoming's HB 102 and South Carolina's H 3431 make it into their count but not MultiState's.
2. Signature status. New York passed two separate chatbot bills through its legislature in 2026: the Safe by Design Act, folded into the FY27 budget, and S 9051, a standalone kids'-chatbot bill that cleared the Assembly 137-0 and the Senate 60-0 in June. Only the first one has actually been signed. S 9051 is still sitting on Governor Hochul's desk — she has until December 31, 2026 to act on it. I count the first and not the second, because "passed unanimously" and "signed into law" are not the same fact, even when it's tempting to treat them that way.
3. Timing. California's SB 1119 — "Adam's Law" — was signed September 10, 2026, two weeks before I'm writing this and after the Transparency Coalition's mid-year report had already gone to print. It's real, it's enacted, and it isn't in anyone's running total yet.

Net all three out, and the count today is 14 states, 14 measures, all verifiably signed into law. That happens to be the same headline number I started with — I promise that's not because I worked backward from a title I liked. It's what's left after I stopped counting a bill that hasn't been signed and started counting one that has.

The freshest signature: California's "Adam's Law"

SB 1119 was signed into law on September 10, 2026. Per State Senator Steve Padilla's office, it makes California "the first state in the nation to enact a comprehensive regulatory framework governing AI chatbots and children." It's California's second companion-chatbot statute — SB 243, which has been in effect since January 1, 2026, was the first — and Adam's Law substantially expands on it: default age-appropriate protections for every child account unless a platform verifies otherwise, limited memory and session caps for minors, mandatory crisis-response protocols, and independent third-party safety audits.

Core requirements take effect July 1, 2027. Those audits don't start until January 1, 2029.

Here's what that record looks like once it's in Matterwire's own tracker:

The January 1, 2027 cluster — now six states, not four

I almost published this post with four states in the January 1, 2027 cluster, because that's what fit neatly into the version of this table I drafted from a secondary source. Then I checked each bill's own chatbot-specific effective date directly, and found two more: Connecticut's SB 5 bundles companion-chatbot rules into a much larger omnibus AI law, most of which takes effect October 1, 2026 — but the chatbot provisions specifically are carved out to January 1, 2027. Colorado's HB 1263 has the same structure: the law itself has been on the books since mid-August, but the actual operator requirements — age estimation, disclosures, self-harm protocols — don't bind until January 1, 2027 either. Both would have been easy to miss by reading only a bill's general effective date instead of its chatbot-specific one.

That's all 14. Six of them — Oregon, Washington, Georgia, Rhode Island, Connecticut, and Colorado — hit the exact same date. If you're building or buying a product with anything resembling a companion or emotionally-responsive chatbot interface, January 1, 2027 is the date that matters most, regardless of which state count you started from.

One thing worth flagging directly: at least one widely cited legal-industry tracker cross-lists Washington's bill under Oregon's number — "HB 2225 / SB 1546" as if they were the same measure. They aren't. HB 2225 is Washington's bill; SB 1546 is Oregon's. Different states, different legislatures, different bill text. It's a small, easy mistake to make when you're aggregating secondary sources instead of checking the primary bill text against a source URL and a timestamp — which is the entire argument for verifying at the record level instead of trusting a summary of a summary.

Why this category is moving so fast

State AI legislation isn't a slow-moving category anymore. MultiState reported 1,200 AI-related bills introduced across all 50 states in 2025; by March 2026, lawmakers in 45 states had already introduced 1,561 more. The Transparency Coalition's own count shows the enactment rate accelerating alongside it: 84 new AI-related laws passed or enacted across 27 states in just the first half of 2026, ahead of the 73 enacted across all of 2025.

Companion chatbots are the sharpest edge of that trend precisely because the underlying harm — a minor forming an unsupervised, emotionally significant relationship with an AI system — is concrete enough that legislatures on both sides of the aisle have moved on it inside a single session. Georgia and Idaho passed theirs through Republican-held legislatures; Washington and New York through Democratic ones. I can't think of many other AI policy questions right now where that's true.

What's next: the GUARD Act

The federal GUARD Act (S. 3062), sponsored by Senators Hawley and Blumenthal, passed the Senate Judiciary Committee unanimously on April 30, 2026. As of today, nearly five months later, it still has no scheduled floor vote. Until it passes — if it passes — the 14-state patchwork above is the operative compliance picture for any product with a chatbot-style interaction surface. A single federal standard would simplify this considerably. It isn't one yet, and I wouldn't plan a compliance timeline around it becoming one this year. It's exactly the kind of jurisdiction-by-jurisdiction patchwork Matterwire's coverage exists to track in one place instead of five separate ones.

The pattern behind this post

This is the first in a series I'm planning to keep running on the same structure: a jurisdiction count I've personally re-verified, a disclosed methodology for how I got there, and a deadline that turns "which states regulate this" into "what do I need to do, and by when." The next two on my list: a rule-by-rule comparison of the EU's newly enforceable AI Act transparency regime against the UK and US, and a single-date roundup of everything — AI, privacy, pay transparency — taking effect on January 1, 2027 across all five jurisdictions Matterwire's platform tracks.

If you'd rather not re-derive this table yourself every time a legislature moves, that's the actual product. Book a Matterwire demo and I'll show you the live version of the tracker this post came from.