One state now requires the big labs to register
New York opened a registration portal on November 10 for AI developers with over $500 million in annual revenue. The RAISE Act takes effect January 1, 2027. Covered labs—OpenAI, Anthropic, Google, Meta—must register, publish safety frameworks on their websites, report critical safety incidents within 72 hours, and issue public transparency reports. Fines go up to $1 million for a first violation, $3 million for repeats.
This doesn't change what you do with AI today
The registry targets the labs that build the models, not the businesses that use them. If you use ChatGPT, Claude, or Gemini in your business, you don't register and you don't file. The law asks the big companies to show their work, not you to report yours. Nothing about this announcement changes what tasks you can hand to AI this week.
Watch what the labs publish, and what they call an incident
The labs must publish their safety frameworks—the rules they follow before releasing a new model—starting in January. Read them when they come out. If a lab reports a critical safety incident, check what happened and whether it touches the tasks you hand to AI. If your AI vendor won't say how it decides what's safe to release, that's a reason to ask why, or to check who else you could use instead.