Will the U.S. create a new federal agency to regulate AI by the end of 2024?
➕
Plus
25
Ṁ1789
Jan 3
13%
chance

This option for regulation of artificial intelligence has been suggested by various lawmakers, as well as Sam Altman.

To count, the organization will have to have regulation of artificial intelligence as a major portion of its mission/mandate, but it doesn't have to be literally the name of the agency. Since this determination is subjective, I will not bet in this market.

The organization created will qualify as a "federal agency" if it is structured similarly to any of the FDA, NRC, CFTC, SEC, etc. It must have full time federal employees who are not "on loan". So something like the EMP Commission (which gathered various experts temporarily to study the topic) would not count.

This will resolve YES if legislation has been passed and signed to create the organization, even if it hasn't gone into effect yet.

Get
Ṁ1,000
and
S1.00
Sort by:

@ErickBall "FDA, NRC, CFTC, SEC, etc." seems like a really small and unusual subset of all federal agencies. Other than the FDA, all of the agencies you listed here are independent regulatory agencies, which make up only a small portion of all federal agencies.

Can you tell me whether the following agencies count as "federal agencies" for the purposes of this market (if they were new, which they are not)?

Also, does "legislation" have to be "passed and signed to create the organization", or can it be an agency established by executive or administrative action? For example, the National Protection and Programs Directorate was established by administrative action in 2007, and later became known as the Cybersecurity and Infrastructure Security Agency in 2018 (see history here). Did the former National Protection and Programs Directorate count as an agency in 2007, or did it only become one for the purposes of this market in 2018?

Another example: the Board of Immigration Appeals is pretty clearly a federal agency, but it was created by the Attorney General without specific Congressional action.

@octothorpe Thanks for pointing out the ambiguity! As you say, I was imagining an independent regulatory agency to be the most plausible approach, but it's certainly not the only possibility.

I am inclined to interpret "agency" somewhat generously, for instance if it's created as a part of the Executive Office of the President, or under the Department of Commerce or something, rather than as an independent regulatory agency--for example, NIST and NOAA are clearly federal agencies. The key factor is whether it has the authority to regulate AI generally (rather than, for example, drafting a "national AI strategy" or creating policies for how the government should use AI or something). The Office of the National Cyber Director, for instance, is purely advisory, but if it had regulatory authority then it could count.

That said, since the Wikipedia article you linked to includes everything under the sun, I do want to specifically rule out some possibilities:

  • Courts, or anything under the judicial branch will not count (since they interpret laws and regulations, but don't create them)

  • Advisory boards or councils will not count

  • AI-related offices established under existing agencies for the purposes of improving or directing the way AI affects or is used by that agency, or by the US government, will not count

  • AI-related offices under existing agencies that use only their existing regulatory authority will not count, e.g. if there is an EPA Artificial Intelligence Branch that examines the environmental impact of AI in order to help enforce compliance of AI companies with the Clean Air Act, that would not qualify as a new agency.

    It had not occurred to me that an agency could be created by executive authority without legislation, but, consistent with the original wording of the question, I see no reason that shouldn't still count as created by "the U.S."

Wasn’t the last one DHS? That was during an incredibly rare moment of national unity, and even base rate the current probability is way too high