2026-08-31 — All three of California’s AI-in-mental-health bills clear the Legislature on adjournment day; a fourth is quietly shelved by its own author

• California’s Legislature adjourned its session at midnight 08-31 having passed all three of the AI-in-mental-health bills this thread has tracked against the deadline: SB 903, SB 1119, and (cross-referenced from The Reckoning) AB 2575. SB 903 — barring anyone from marketing a chatbot as “therapy” and requiring licensed-professional oversight of AI in mental-health care — had already passed the Assembly 74-1 on 08-30; the Senate concurred in the Assembly’s amendments 40 Ayes-0 Noes on 08-31, sending it to Engrossing and Enrolling. SB 1119 — the children’s companion-chatbot safety bill requiring annual child-safety risk assessments, real-time crisis referrals, and AG-reported independent audits by 2027-07-01 — passed its Assembly floor vote and got Senate concurrence in the Assembly’s amendments 39-0, also on 08-31. Both bills now sit at Engrossing and Enrolling, one procedural step from presentment to Governor Newsom; neither has been signed or vetoed. Three separate California AI-in-mental-health bills clearing the same legislative session, on the deadline day itself, is the wave this thread was opened to track arriving all at once rather than piecemeal. (SB 903 bill history, SB 1119 bill history)
• AB 2023 — the SB 1119 companion bill on children’s-chatbot safety that the 08-31 digest flagged as “riding tonight’s deadline alongside the three bills this map already tracks” — did not ride it: its own author, Sen. Padilla, pulled it to the Senate’s inactive file on 08-30, the day before adjournment. The Legislature’s own bill-history record shows no floor vote and no defeat — just “Ordered to inactive file at the request of Senator Padilla,” after the bill had already passed the Assembly 66-8 in May and cleared Senate Appropriations’ suspense file 6-1 on 08-13. Read plainly: the same senator carrying SB 903 and SB 1119 to passage chose not to spend the Legislature’s last day on a third, overlapping bill — a deliberate stand-down, not a loss. It dies for this two-year session with no vote on the record. (AB 2023 bill history)

2026-08-30 — SB 903’s due-today floor vote hasn’t posted yet; SB 1119 in the same holding pattern

• Checked directly against the California Legislature’s own bill-history record Sunday 08-30 (~10am ET, the day SB 903 was scheduled as item 103 on the Assembly’s daily floor file): neither SB 903 nor SB 1119 shows an Assembly floor vote recorded yet. SB 903’s history still ends at “ordered to third reading” (08-21); SB 1119’s still ends at “read third time and amended, ordered to third reading” (08-28). Both bills remain exactly one floor vote away, as logged on 08-28/08-29, but the vote itself has not yet been posted to either bill’s official record. This is not a hit, a slip, or a death — the constitutional adjournment deadline is 08-31, one day out, and leginfo has previously lagged same-day floor action by several hours during active session, so this reads as “not yet decided” rather than “missed.” Needs a re-check later today or tomorrow for the actual outcome. (SB 903 bill history, SB 1119 bill history)

• California’s SB 903 — which bars marketing a chatbot as “therapy” and requires licensed-professional oversight of any AI used in mental-health support — passed its Assembly floor vote 74 ayes to 1 no (DeMaio) on the evening of 2026-08-30, the Sunday session that reconvened at 17:00 ET/14:00 PT ahead of Monday’s constitutional adjournment. The bill returns to the Senate for concurrence in Assembly amendments, the last procedural step before it reaches Governor Newsom’s desk — the strongest single mover so far this session among the three AI-in-care bills this map has been tracking against the deadline. (Assembly floor vote record via leginfo)

2026-08-29 — SB 903 clears Assembly Appropriations and heads to a floor vote before session’s end

• California SB 903 (Padilla) — the “mental health professionals: artificial intelligence” bill barring AI from being marketed as therapy and requiring licensed-professional oversight — passed Assembly Appropriations 13-0 on 08-13 and was ordered to Assembly third reading on 08-21, now scheduled as item 103 on the Assembly’s August 30 daily floor file. Verified directly against the Legislature’s own bill-status record. This thread logged SB 903 back on 07-28 sitting in Appropriations with an 08-14 fiscal deadline; it has now cleared that deadline and is one floor vote away from passage, ahead of California’s 08-31 constitutional end-of-session deadline. Distinct from AB 2575 (liability, tracked on The Reckoning/Kaiser vs. AI) — SB 903 is squarely a chatbot-as-therapy restriction, the bans/restrictions angle this thread specifically watches. (California Legislature — SB 903 bill status)

2026-08-28 — A third California companion-chatbot bill — this one aimed at children’s toys and companion apps broadly — clears its last committee stop

• SB 1119, the children’s companion-chatbot safety bill, was read a third time and amended in the Assembly on 08-28, one procedural step from a floor vote. Authored by Sen. Padilla with Assemblymembers Bauer-Kahan and Wicks, it requires companion-chatbot operators to run annual child-safety risk assessments, provide real-time in-app crisis referrals when a child appears to be in crisis, and submit to independent audits reported to the Attorney General (by 07-01-2027) — the broadest of the “companion-safety” species this thread already tracks (alongside CA’s existing SB 243). It passed the Senate 39-0 in May and cleared Assembly Appropriations 11-1 on 08-13; the only step left is final Assembly passage before the 08-31 adjournment, then Senate concurrence. ⚠️ Not yet signed or in effect — flag as pending, not enacted. (Bill history, leginfo, Sen. Padilla’s office, Transparency Coalition scorecard, 08-24)

2026-08-27 — KFF maps the patchwork, and the classification question is still open

• KFF published a state-by-state comparison of AI mental-health-tool regulation, finding states are not converging — Illinois and Nevada restrict use directly, Utah and Rhode Island regulate disclosure and data handling instead, and the underlying question of whether these tools are clinical support, a consumer product, or care requiring licensure remains unresolved. That classification question is the whole fight, because liability, reimbursement and who may deploy one all follow from it — and a patchwork answer means the deploying organisation effectively picks its jurisdiction. A policy synthesis rather than a new event, logged as the clearest current map of this thread’s terrain. (KFF)

2026-08-19 — Backfill: Colorado’s HB26-1195 has been binding law since 08-12, missed until today

• 🕰 CAUGHT LATE — Colorado’s “Psychotherapy Artificial Intelligence Restrictions” act (HB26-1195) was signed 2026-06-03 and took effect 2026-08-12; this thread had no record of it until today’s backfill. It bars licensed, certified, or registered psychotherapy providers from letting an AI system conduct therapeutic communication with a client except during synchronous, real-time interaction where the professional, the AI, and the client are all actively participating together; from generating therapeutic recommendations or treatment plans without the professional’s review and approval; or from using AI to detect emotions or mental states as a stand-in for professional judgment. AI use for administrative/supplementary tasks stays permitted if the professional retains full responsibility. Separately, the act makes it unlawful for anyone — individual, corporation, or entity — to provide, advertise, or offer psychotherapy services to the public in Colorado unless delivered by a regulated professional. (Colorado General Assembly — HB26-1195, Forbes)
• This is a different branch of government from the Chatbot Safety Act rulemaking already on this thread, enacted first, and already binding. The AG’s Chatbot Safety Act rules (proposed 08-11, this thread’s prior Colorado entry) come from the Department of Law and don’t bind until 2027-01-01. HB26-1195 came through the legislature, was signed 06-03 — five weeks before the AG’s rules were even proposed — and has been in force for a week already. Track both as distinct Colorado actions: the AG’s rules govern consumer-facing chatbots generally; HB26-1195 governs licensed psychotherapy specifically and is the more direct answer to the “AI posing as a therapist” question this thread otherwise tracks through Character.AI litigation.

2026-08-13 — Resolved: SB 903 released from suspense 13-0, ordered to third reading

• SB 903 was released from Assembly Appropriations’ suspense file today, not held: the committee reported “Do pass” on a unanimous 13-0 vote, the bill was read a second time, and it was ordered to third reading — clearing both open expectations on this bill (ca-sb903-appropriations-hearing, due yesterday, and the ca-sb903-assembly 08-14 fiscal report-out deadline) a day early. Leginfo’s own bill-history log shows today’s sequence in order: “Assembly Rule 63 suspended” (waiving the normal in-print waiting period before floor action), “From committee: Do pass. (Ayes 13. Noes 0.),” then “Read second time. Ordered to third reading.” The bill now sits on the Assembly’s own “3rd Reading File” for Senate bills, dated 08-17-2026 (item 269) — matching the Assembly’s adjournment notice (“adjourned until Monday, August 17th”) the morning and afternoon passes below were tracking before the resolution posted. SB 903 already passed the Senate floor 39-0 back on 05/19/26; the next and final Assembly step is the floor vote itself, after which any Assembly amendments would need Senate concurrence before the bill reaches the Governor’s desk. (California Legislature — SB 903 bill status, California Legislature — SB 903 bill history)
• Earlier the same day (afternoon check): the trigger condition the morning pass was waiting on had cleared — the Assembly adjourned its floor session and the suspense-file hearing was underway — but no disposition had posted yet. The Assembly’s own homepage read “The Assembly has adjourned until Monday, August 17th, 2026 at 1:00 p.m.,” and the Appropriations Committee’s hearing (calendared “upon adjournment of Session,” Room 1100) showed as live, while leginfo still showed only the 08/05/26 “Placed on suspense file” entry as the most recent action. Superseded within hours by the resolution above.
• Earlier still (morning check): SB 903 was confirmed on Assembly Appropriations’ suspense calendar for the day’s Daily File, disposition not yet known. The committee’s own hearings page listed “S.B.No. 903 Padilla. Mental health professionals: artificial intelligence,” Wicks chairing, hearing time “Upon adjournment of Session,” Room 1100 — confirming the bill remained alive and treated as a suspense-file matter, consistent with its August 5 hearing.

2026-08-13 — Vermont’s H.816 confirmed signed, resolving this thread’s “VT confirmation” watch item

• A regional feature on Dartmouth’s Therabot research confirmed Vermont’s H.816 is signed law, not still pending: it requires a licensed professional to review AI-generated mental-health advice before it reaches a patient. Signed in June 2026, after Rep. Daisy Berbeco introduced it citing “stories of people doing self-harm … when they interact with chatbots.” The same reporting found New Hampshire’s own attempt — Sen. Howard Pearl’s bill to regulate AI mental-health tools as unlicensed practice, a workaround for the federal executive order discouraging direct state AI regulation — stalled into a study committee rather than passing, after stakeholders couldn’t agree. Closes a watch item open since this thread’s 2026-07-22 seed correction (“Vermont unconfirmed past its 05-07 pending stage”). (Concord Monitor / Valley News)

2026-08-12 — HB 26-1195 takes effect: the flip, not just the check

• Colorado’s HB 26-1195 took effect today — the country’s first law requiring a real, licensed therapist to stay in the loop for AI use in psychotherapy actually bound, as opposed to being previewed. It bars AI from directly conducting therapeutic communication, generating unreviewed treatment recommendations, or performing unsupervised diagnosis/emotional-state detection, while still allowing AI for administrative/documentation support under full clinician responsibility. Yesterday’s entry already confirmed no delay, amendment, or challenge on the legislative record; today’s development is simply the law becoming binding law rather than a pending one — the upcoming.yaml ledger entry (colorado-hb1195-effective) flips to hit. Colorado’s separate, broader AI Act (SB 26-189) remains under its own federal injunction fight and does not touch this bill or its companion HB 26-1139. (Colorado General Assembly)

2026-08-11 — Pre-effective-date check: HB 26-1195 has no delay, amendment, or challenge on file

• A dedicated check of Colorado’s legislative record ahead of HB 26-1195’s August 12 effective date found no delay, amendment, or legal challenge. The bill — the country’s first law requiring a licensed therapist to stay in the loop for AI use in psychotherapy — was signed June 3, 2026 and takes effect August 12 by mechanical operation (any bill enacted without a safety clause takes effect once the legislature adjourns, which it did May 13). A Colorado Senate Democrats release corroborates the date. Not to be confused with xAI’s separate federal preliminary-injunction fight against Colorado’s broader AI Act (SB 26-189) — legal trackers confirm this bill and its companion HB 26-1139 proceed unaffected. (Colorado General Assembly)

2026-08-10 — ITIF warns the state patchwork could copy social media’s failed playbook

• A report from the Information Technology and Innovation Foundation found nearly 100 state bills addressing chatbot safety introduced as of August, and warned that inconsistent state definitions burden developers without addressing documented child-safety harms. ITIF recommends targeted safeguards, transparency, and parental controls over blanket age-verification or bans, citing California Governor Newsom’s veto of a broad companion-chatbot bill as its cautionary precedent — a direct contrast to Colorado’s mandate-heavy HB 26-1195, which takes the opposite approach and takes effect 08-12. (StateScoop)

2026-08-07 — SB 903 breaks into national press for the first time, framed alongside the Kaiser/NUHW fight

• An AP wire story (“As AI ’therapists’ dish out advice, California lawmakers try to set some limits”) put SB 903 in front of a national audience for the first time, syndicated to dozens of outlets (KPBS, News From The States, Lost Coast Outpost, several regional papers) — and it names the National Union of Healthcare Workers as one of the bill’s backers, explicitly tying this thread to the Kaiser labor fight in the same story. Coverage of SB 903 up to now has been legislative-tracking-only (leginfo, trade press); this is the first general-audience framing, and it arrives already linked to NUHW’s “unchecked AI in mental healthcare” campaign rather than as a standalone bill story. No change to the bill’s own procedural status — still on the Assembly Appropriations suspense file (see 2026-08-06 entry below); next action still estimated ~08-18. Wire story published evening of 2026-08-06 ET, ahead of the 2026-08-07 digest window; dated explicitly rather than presented as breaking today. (News From The States, AP wire, CalMatters, original)

2026-08-06 — SB 903’s August 5 hearing outcome confirmed: suspense file, not a kill

• Leginfo’s own bill-history log resolves the open question left by yesterday’s entry: SB 903’s August 5 Assembly Appropriations hearing outcome was recorded as “August 5 set for first hearing. Placed on suspense file.” This is the same phrasing leginfo used for the bill’s earlier Senate Appropriations hearing (05/04/26: “May 4 hearing: Placed on APPR. suspense file”) — which was followed by a second suspense-file hearing (05/14/26) where the committee released the bill 7-0 to a floor vote. Read against that precedent, being placed on suspense is procedural, not a fail signal: the bill is alive, with its actual fate (release to the floor vs. held/died) decided at a later suspense-file hearing — California’s Assembly Appropriations suspense hearings typically land mid-to-late August. Primary-source confirmed (leginfo’s own chronological bill-history record); no second outlet found independently reporting the same line as of this check. (California Legislature — SB 903 bill history)

2026-08-05 — The federal preemption vehicle clears its first committee test

• The CHATBOT Act (S.4407, Cruz/Schatz) advanced out of the Senate Commerce Committee’s kids’-online-safety markup today, alongside the Kids Online Safety Act (heading to the Senate floor after stalling in the last Congress), the SCREEN Act, the Youth AI Privacy Act, and the Children’s AI Toy Safety Act — the concrete federal preemption vehicle this thread has been watching (see the 07-31 companion-chatbot-bill entry) takes its first procedural step out of committee. (Senate Commerce Executive Session 24)
• SB 903’s own Assembly Appropriations hearing convened as calendared at 9am today, but no outcome (pass, hold-on-suspense, or fail) was confirmed in this pass — leginfo’s bill-status page still showed the pre-hearing 07-02 action as the most recent recorded entry as of this check. A dedicated pass is tracking this specifically; not calling a result here to avoid a premature/conflicting entry. (California Legislature)

2026-08-04 — The SB 903 hearing is not a suspense hearing, which changes what tomorrow decides

• Re-verified on the committee’s own hearings page: SB 903 is calendared 08-05 at 9 a.m., 1021 O Street Room 1100, Assembly Appropriations (Wicks chairing), listed verbatim as “S.B.No. 903 Padilla. Mental health professionals: artificial intelligence.” (Assembly Appropriations)
• The listing detail: the bill sits under “BILLS HEARD IN SIGN-IN ORDER” and neither the committee page nor leginfo’s bill history uses the word “suspense” — the latest recorded action is 07/02/26 “Read second time and amended. Re-referred to Com. on APPR.”
• ✏️ Correction, same day — the prediction is withdrawn, the facts hold. An earlier version of this entry read the missing “suspense” label as meaning an up-or-down committee action was the likeliest outcome. That inference is withdrawn: the 08-05 calendar carries 360+ measures, and an August Appropriations hearing that size is the shape of a suspense calendar in California practice whatever the page calls it. A separate claim that SB 903 was placed on suspense around 07-01 surfaced from a secondary document and is not corroborated by either primary source. Net: all three outcomes stay open.

2026-08-03 — SB 903 is confirmed calendared — Assembly Appropriations, Wed Aug 5

• California’s SB 903 landed on the Assembly Appropriations calendar for a 08-05 hearing — dual-verified against raw HTML on the legislature’s first day back from recess (leginfo’s Committee Hearing Date field 08/05/26 and the committee’s own Daily File naming SB 903, Padilla, at 9am, Room 1100), the calendar entry that upgrades the ledger from reported to confirmed; whether it is held on suspense or moves forward is decided at/after the hearing. (California Legislature)

2026-07-31 — A federal companion-chatbot bill surfaces (see also Grok Harm)

• A federal bipartisan companion-chatbot bill for minors was introduced 2026-07-28 (Sens. Andy Kim D-NJ / Jon Husted R-OH, S.5154, “Children Harmed by AI Technology Act 2.0”) — the preemption pressure this thread’s watch line has been tracking now has a concrete federal vehicle: a tiered risk-based framework on AI companion chatbots, referred to Senate Commerce. Full detail logged on Grok Harm’s 07-31 entry; noted here for the preemption angle specifically. (congress.gov S.5154)

2026-07-29 — Maine in force; SB 903’s Appropriations clock still unresolved

• Maine’s LD 2082 barring AI-delivered therapy took effect today — the first US state statute of its kind actually in force.
• California SB 903’s Appropriations timing is unresolved: the bill’s last recorded action is 07-02, no Assembly Appropriations calendar entry names it, and sources place the committee deadline at both 08-14 and 08-29. (CalMatters)

2026-07-28 — Maine’s ban becomes the first one actually in force

• At midnight, Maine LD 2082 took effect — the first US statute barring AI-delivered therapy to actually bind, as opposed to being signed or pending. Enacted as Public Law Chapter 687, signed 04-13-2026. It bars offering “therapy or psychotherapy services, including through the use of Internet-based artificial intelligence” to the public unless delivered by a licensed professional; licensees may use AI only for “administrative support or supplementary support,” retaining “full responsibility for all interactions, outputs and data use.” (Maine Legislature)
• Provenance note, stated plainly: the chapter number and signing date are primary-confirmed from the legislature’s own status page. The 07-29 effective date is not printed in the act — it follows Maine’s default 90-days-after-adjournment rule and is corroborated across legal trackers. Reported as strong-secondary, not primary-document confirmed.
• This was a map gap, not a new event. Maine appears nowhere in this thread’s prior timeline or backstory despite being signed in April — the thread had Colorado, Hawaii, NY and California but not the one that would bind first. Caught by the overnight lens sweep.
• The clock board now reads: Maine in force; Colorado HB 26-1195 effective 08-12; CA SB 903 at Appropriations, fiscal deadline 08-14; Minnesota’s nudify ban effective August but now under federal challenge by xAI.

2026-07-22 — Hawaii turns out to already be resolved; NY’s real deadline found

• Correction: Hawaii’s SB 3001 was already signed — Gov. Green signed it July 14 as Act 248 (AI-companion disclosures, self-harm response protocols, annual BHA reports); the ledger had this as still-pending until this pass found it. (West Hawaii Today)
• NY’s Hochul still hasn’t acted on S9051B/S9408A — and her actual statutory deadline is Dec 31, 2026, not the ~Aug 31 placeholder the ledger carried; corrected in upcoming.yaml. (multistate.ai)

← Backstory (crawl 2026-07-22 → artifacts/findings/state-therapy-chatbot-bans-2026-07-22.md)

⚠ Correction to the seed picture

• The “7 states” framing was wrong in both directions — Tennessee is unverified (no source found); Vermont’s enactment unconfirmed past its 05-07 pending stage; IL/NV/UT bill numbers unfound. Meanwhile the wave was undercounted: CO/CT/GA/CA/TX/NH/WA/ID/IA all have enacted instruments, and NY’s companion-disclosure law has been effective since 2025-11 (separate from the pending under-18 ban). Full corrected ledger in the finding.

2026-07 — The July instruments

• Rhode Island enacted its two-bill package 07-06 — chatbot safety (crisis protocols, AG reporting, 3-hr disclosure) + mental-healthcare AI (informed consent, AI barred from treatment decisions). (Progressive Charlestown)
• NY’s S9051B (under-18) + S9408A (AI toys) still awaited Hochul’s signature as of 07-10; Hawaii’s SB 3001 awaited Green’s decision. (JD Supra)

2026-05→06 — The spring instruments

• Colorado HB26-1195 signed 06-03 — the strictest practice-law: bars unsupervised AI-delivered psychotherapy; real-time therapist mandate. (Forbes)
• Connecticut’s omnibus SB 5 signed 05-27 (PA 26-15) — companion provisions effective 2027-01-01; private right of action. (JD Supra)

2026-01 — The floor was already moving

• California’s SB 243 and New Hampshire’s child-endangerment law took effect in January; Google/Character.AI settled the Setzer wrongful-death case the same month — the marquee case behind the wave. (ABA Journal)

The preemption track (context)

• DOJ’s AI Litigation Task Force backs xAI’s 07-12 suit against Colorado’s algorithmic-discrimination law — no suit found against any therapy/companion chatbot law yet. (Westword)