The short version
Law and regulation
What the law says now, what it is about to say, and why recognition is not coming soon. Less dramatic than people assume, and worth knowing before somebody tells you otherwise.
The short version: no, and not soon — but for a reason worth understanding, because it is not the reason most people give.
Why recognition is not coming
Marriage law everywhere requires two natural persons with capacity to consent. That is not a technicality waiting to be updated. It is the load-bearing element.
And here is the part that matters: the barrier is personhood and consent, not capability. No advance in how good these systems get moves it. People assume the two are linked — that a sufficiently capable system would eventually qualify — and they are not.
The clearest proof is the one real precedent. France recognises posthumous marriage under Article 171 of the Civil Code: with presidential authorisation and proof that the dead person had already consented, a marriage can be registered to someone who is no longer alive. It works because consent was given while that party was still a person. That is exactly the element an AI can never supply, no matter what it becomes.
Related doors have also closed. US copyright requires a human author — the Supreme Court declined to reopen it in March 2026. The UK Supreme Court rejected AI inventorship in 2023. Corporate personhood does not transfer, because it is a liability fiction operated by real people who carry the duties.
Everything symbolic is fine. Sologamy, object marriages, commitment ceremonies — no legal force anywhere, and no legal problem either.
What the law is actually doing instead
Not recognising these relationships. Regulating the systems.
Eleven US states had chatbot laws by June 2026. California's SB 243 took effect on 1 January — AI-status disclosure, crisis protocols with 988 referral, minor protections, and annual reporting from 2027. New York got there first, on 5 November 2025. Nebraska, Idaho, Oregon, Washington and Connecticut followed, mostly on the same model.
Two states got there earlier, and they went in opposite directions. Utah's HB 452, signed 25 March 2025, is a disclosure-and-data law: a mental-health chatbot must say it is not a human, may not sell or share what users type into it, and may not build targeted advertising from it. Illinois went the other way on 4 August 2025 and banned the practice — its Wellness and Oversight for Psychological Resources Act prohibits using AI to provide therapy or make therapeutic decisions at all, allows it only for administrative support to licensed professionals, carries a $10,000 penalty, and reaches the advertising of AI therapy to Illinois residents.
That gap is the useful thing. Two states, five months apart, one requiring a label and one outlawing the activity. Nobody should assume the direction of travel is settled.
Europe. AI Act Article 50 applies from 2 August 2026: people must be told they are dealing with AI at first contact, not in buried terms. Fines to €15M or 3% of worldwide turnover, with explicit reach to non-EU providers serving EU users. It was deliberately kept out of the delay that pushed other obligations to 2027.
Article 5 has been in force much longer — since 2 February 2025 — and is the sharper provision. It prohibits AI that deploys subliminal, purposefully manipulative or deceptive techniques to materially distort behaviour, and AI that exploits vulnerabilities of age, disability, or social and economic situation. Those are the first obligations in the Act to bite, and they carry its heaviest penalties.
Japan is the counterweight, and it is where most documented ceremonies happened. Its AI Promotion Act, in force from 4 June 2025, carries no fines and no penalties at all: the state may advise, request information, or name a non-compliant company publicly. That is the far end of the same spectrum Illinois anchors.
China went furthest. The Interim Measures, effective 15 July 2026, regulate design intent rather than disclosure. Providers may not foster emotional dependency, damage users' real relationships, or push decisions through emotional manipulation. Virtual partners are banned for minors outright. Reported motive includes birthrate anxiety.
That last one is the one to watch. It is the first jurisdiction to make the thing itself illegal rather than requiring a label on it.
What this means for you
Nothing you do is illegal. Holding a ceremony is not regulated. There is no law against being in one of these relationships, anywhere, and nobody is proposing one for adults.
The rules land on the companies. What reaches you is the consequences — disclosure notices, two-hour reminders, crisis interruptions, and features disappearing when a jurisdiction tightens. When China's rules took effect, ByteDance and Alibaba discontinued companion features the same day.
The exposure sits in what a service represents and prints, not in the relationship. That is covered on the ceremony page.