VERIFIED 29 JUL 2026 · Three regimes · One technology
The EU wrote a single rulebook. The US is racing with no federal statute. China makes you file before you launch. Step into the arena and see whose rules govern your AI, use case by use case, with every claim carrying its source.
Exhibit 01 · The contenders
Each regime answers the same question in a different key: who bears the risk, who checks the work, and who has to ask permission first.
The rulebook · The Pyramid
One law. Rights first.
The world's first horizontal AI law sorts every system into a four-tier risk pyramid: the more it can affect lives and rights, the heavier the rules.
€35M / 7%
Top fine
Bigger than GDPR, for prohibited practices.
The patchwork · The Patchwork
No federal law. Race to win.
No comprehensive federal statute. Executive orders chase innovation while states, cities and sectoral agencies write the only binding rules.
99 to 1
Senate vote
Killing the 10 year state law moratorium, July 2025.
The registry · The Registry
File before you launch.
Vertical, state led rules, one per technology, backed by a public filing system: algorithms and generative AI services register with the CAC before launch.
868
GenAI filings
Services filed with the CAC by April 2026.
Next on the docket
On that date the EU's Article 50 transparency duties switch on: chatbots must say they are AI and deepfakes must be labeled. The Commission's enforcement powers over general purpose AI providers activate the same day, and so does California's SB 942 content provenance mandate. Brussels and Sacramento, one deadline.
What goes live
Exhibit 02 · The simulator
Six AI systems, scored against each regime with the reasoning and the citations behind every verdict.
CV screening & candidate ranking
general-purpose customer service
AI-driven diagnostic device
synthetic media tool
automated creditworthiness
real-time biometric ID in public
The scoreboard so far
€35M
EU top fine
Or 7% of worldwide turnover for prohibited practices, the toughest penalty in EU digital law.
~85%
EU systems untouched
Commission estimate of AI systems in the minimal risk tier with no new duties.
109
US state AI laws in H1 2026
Enacted despite a federal preemption push from the White House.
1,247+
FDA authorized AI devices
The most concrete algorithmic oversight in America, roughly 76 to 80% of it in radiology.
5,672
Deep synthesis algorithms filed
14 public batches on the CAC registry as of January 2026, with 32 deregistered in 2025.
RMB 200k
CN companion AI fine ceiling
Under the world's first AI companion rulebook, in force 15 July 2026.
Exhibit 03 · AI literacy
Nine claims you can make in a meeting and defend with the primary source.
The fines are bigger than the ones in GDPR.
You can be fined without setting foot in Europe.
The same AI can be illegal in one room and legal in the next.
The first federal US law regulating AI content is about deepfake pornography.
America's first comprehensive state AI law never took effect. Its own state repealed it.
The most consequential AI regulator in the US is the FDA.
China publishes a public registry of every approved AI service.
More than 5,600 deep synthesis algorithms are on file, and filings can be cancelled.
AI companion apps were forced to break up with their users, by law.
Exhibit 04 · Method
Every verdict, date and fine here is drawn from primary legal texts and specialist trackers, cross checked as of 29 July 2026. Verdict labels are editorial summaries of the cited instruments: this is an educational simulator, not legal advice. Where a claim rests on a single source or is contested, the citation panel says so.
Verdict key
Status key
A status says whether an instrument is actually law right now. The word carries the meaning, so the colour is reinforcement and never the only signal.
Official and primary
Law firm and specialist analyses
Official and primary
Primary, official and translations
Law firms, trackers and analysis
The arena shows you the map. GAGE programs put you on the ground in the jurisdiction you actually have to answer to.
Study aid, not legal advice. The official texts are always authoritative: start with the sources above, or read the AI Act itself in the EU AI Act Explorer. The United States corner of this arena is a patchwork of state laws: see all of it, state by state, on the 50-State AI Law Atlas.