AI regulation has gone properly global and entered the legal big leagues. If your AI is in production, it’s more than likely in scope.
So what?
If you’re building or deploying AI in regulated environments, the question is no longer “Is there regulation?”
It’s “Which regime applies, and can you prove control?”
Builders can become “accidental providers” the moment they ship AI functionality.
Deployers still need governance artefacts, monitoring, and incident readiness.
Agentic systems raise the bar again: autonomy means permissions, audit trails, and containment become table stakes.
If you thought the G20 OTC derivatives reforms were chaotic, AI regulation is a different sport. More jurisdictions, faster cycles, and far less agreement on the rulebook.
The US is doing it state-by-state: California’s SB 53 (TFAIA), and Texas HB 149 (TRAIGA), effective 1 Jan 2026.
Add federal politics into the mix and you get the usual outcome: uncertainty and likely litigation, rather than a neat, unified rulebook.
This isn’t “EU vs Silicon Valley” anymore if it ever was. It’s a global patchwork. And it applies to model builders and deployers.