A law guarantees a right to a human doctor or teacher

Where things stand (16 Oct 2026)

As far as I can find, no national law in a G20 country or the EU qualifies yet. The question excludes Italy's Law 132/2025 by name. Italy's two decrees implementing that law and adapting Italian law to the EU AI Act have the force of primary law:

• D.Lgs. 160/2026, published 15 Sept 2026, one day before the opening date. It covers criminal and civil liability and policing.
• D.Lgs. 179/2026, published 8 Oct 2026. It covers national authorities, sanctions, AI training (including mandatory AI content in doctors' continuing education), workplace decisions and fees for professionals.

Neither creates a right to a human clinician or teacher. Sources: Federprivacy, 9 Aug 2026; Quotidiano Sanità, Oct 2026 (https://www.quotidianosanita.it/governo-e-parlamento/ai-dal-23-ottobre-in-vigore-il-nuovo-decreto-controlli-sanzioni-e-formazione-novita-anche-per-lecm/); Ingenio, Oct 2026.

The clearest template is a state law, so it does not count. California's AB 2148, signed 30 June 2026, requires K-12 teachers to be human (EdSource, NY Post, July 2026). Demand is broad: in a Johns Hopkins poll, 79% of Americans want a right to a human in medical care and 74% in education (Futurity, 23 June 2026).

How I read the question. I count a national ban on replacing teachers or therapists with AI, which leaves human delivery as the only lawful option, as a "mandatory duty to provide human-delivered service." California's law and Brazil's bill PL 3003 are examples. I do not count rules that only require human oversight or a doctor's final decision, disclosure, or impersonation bans.

National pipeline (all far from enactment)

• Brazil, the strongest lead.
  • PL 3003/2025 bans replacing basic- and higher-education teachers with AI in "teaching activities." It passed the science and technology committee on 17 Dec 2025. Since May 2026 it has waited for the rapporteur's opinion in the education committee. It still needs the constitution and justice committee, the Senate and presidential sanction. I took its status from the Chamber's open-data API (https://dadosabertos.camara.leg.br/api/v2/proposicoes/2527320/tramitacoes). Its author was re-elected on 4 Oct 2026 (g1), so he can revive it after it is automatically archived at the end of the legislature.
  • PL 3782/2025 would bar fully replacing human professionals in health and education. It is attached to the main AI bill, PL 2338/2023, which is now not expected to be voted until the end of 2026 at the earliest and must then return to the Senate (Mobile Time, 24 Aug 2026).
  • The Senate's health-AI bill PL 931/2026 covers only supervision and transparency (Agência Senado, 14 Oct 2026). The Federal Council of Medicine's resolution 2454/2026 is a professional rule, not legislation.
• Russia. The digital ministry's March 2026 draft would have required organisations to provide services without AI on request, with the government setting the list of cases (Kod.ru, citing Izvestia). The law passed in July 2026 dropped all restrictions (Garant; Pravo.ru). The Duma speaker said lawmakers will revisit it.
• China. A revision of the Teachers Law is due to reach the national legislature's standing committee in 2026 (Jiemian, 11 May 2026). Reported content is about teacher status and pay. A 2025 education ministry guideline bars teachers from using generative AI as a "substitute teaching agent," but it is not law.
• Mexico. Senate committees approved a school cellphone reform of the General Education Law on 12 Oct 2026. It adds AI literacy only, with no clause guaranteeing human teachers.
• US federal, EU and UK. Federal bills are about disclosure and chatbots (Nextgov, July 2026). The EU's KIDS Act proposal is about age-gating, and any right to human contact in the coming Digital Fairness Act would cover customer service, which the question excludes. The UK's NHS AI triage plan offers no statutory right to a human clinician.

Base rate and model

Reference class: national laws in G20 countries or the EU that require humans to deliver care or teaching. There have been none in roughly 3.5 years of mainstream generative AI, while US states have passed several. Ten-plus legislatures could plausibly act, each with a small probability. My rough per-country estimates for 2026–2031: Brazil about 7%, China about 4%, Russia, the US and France about 2–3% each, and roughly 1% each elsewhere. Together these give about 20–27%.

I used quarterly hazards, computed in Python:

• Q4 2026: 0.3% (nothing near enactment)
• 2027: 0.9% per quarter
• 2028: 1.2% per quarter
• 2029: 1.4% per quarter
• 2030: 1.6% per quarter
• 2031: 1.7% per quarter

The rise reflects growing AI substitution and the spread of the policy from US states. The cumulative probability reaches about 24% by the end of 2031.

Arguments against this estimate

• For a higher number: bans on social media for minors spread across G20 countries within about two years once they became popular. A "human teacher" guarantee is similarly simple and is backed by unions and parents.
• For a lower number: governments want AI to ease shortages of doctors and teachers. National laws mostly use oversight wording, which the question excludes. Brazil passes few of the bills its deputies introduce. Taken together, I keep my estimate in the low-to-mid 20s.

Question. By the last day of each quarter, has any G20 country or the European Union enacted a law giving people a right to be treated by a human clinician or taught by a human teacher rather than by an AI system?

Definitions.

• G20 country: Argentina, Australia, Brazil, Canada, China, France, Germany, India, Indonesia, Italy, Japan, Mexico, Russia, Saudi Arabia, South Africa, South Korea, Türkiye, the United Kingdom, or the United States. Only enactment by the national government counts; state, provincial, municipal, and other subnational laws do not. EU membership alone does not make a country eligible.
• Enacted law: a national statute, constitutional amendment, or legislative instrument having the force of primary legislation that has completed the jurisdiction's required enactment steps, including assent or promulgation where required; or a finally adopted EU legislative regulation or directive. A future commencement or transposition date does not prevent qualification. Proposals, passage by only one chamber, delegated administrative regulations, executive orders without the force of primary legislation, and nonbinding declarations do not count.
• Right: a binding entitlement of patients, students, or a defined class of them to receive the relevant service from a human. This may be an express right to request a human alternative or a mandatory duty to provide human-delivered service that leaves providers no option to substitute AI-only service for that class. Enforcement through a regulator or administrative complaint suffices; a private right of action is not required. The right may cover only one clinical specialty, educational level, service, or public-service program; it need not cover both healthcare and education or every resident.
• Human clinician: a natural person legally authorized to deliver clinical care, including a physician, nurse, psychologist, or other licensed clinical professional. Human teacher: a natural person delivering instruction through a school, university, vocational program, or other organized educational service. Diagnosis, treatment, therapy, and substantive instruction count; appointment booking, admissions, insurance determinations, grading alone, and general customer support do not.
• Rather than an AI system: the operative provision expressly addresses AI or automated service delivery and requires human delivery of the covered service, either by default or on request. An AI system is software or a robot that generates clinical care or instruction using machine-based inference. The human may use AI assistance. In-person attendance, AI-free supporting tools, and identical prices or waiting times are not required. Human delivery by telephone, video, or substantive personal written interaction counts; prerecorded human content alone does not.

Resolves YES if a qualifying enacted law establishes the entitlement above in either healthcare or education. An AI opt-out qualifies only if the law also guarantees a human-delivered alternative, not merely permission to forgo the service. A general human-service entitlement covering healthcare or education qualifies if those services fall expressly within its statutory scope.

Exclusions (resolve NO).

• Human oversight, professional accountability, sign-off, or human review or appeal of automated decisions without an entitlement to receive the underlying care or instruction from a human. For example, reserving medical decisions to physicians, as in Article 7(5) of Italy's Law No. 132 of 23 September 2025, is not by itself sufficient.
• Disclosure, consent, privacy, explanation, or AI-refusal rights without a guaranteed human-service alternative; professional licensing requirements or teacher employment protections alone; and AI bans that do not establish a duty to provide human-delivered service.
• Nonbinding guidance, institutional policies, collective bargaining agreements, court judgments without a qualifying legislative enactment, or laws merely authorizing future creation of the right.
• Mere republication, consolidation, or commentary on a previously enacted law. An amendment counts only if it newly establishes or expands a qualifying entitlement.

Resolving source. The enacting jurisdiction's official gazette or official national legislature record, including the authenticated enacted text and evidence that enactment is complete; for the EU, the Official Journal of the European Union, available through EUR-Lex. These are the exclusive resolving sources, rather than the broader Public confirmation convention A2. Apply the Resolution window and Resolution date convention A3 to the first qualifying official publication establishing completed enactment. Apply the once-YES rule even if commencement is later or the law is subsequently repealed or invalidated.