Biden makes serious AI regulation moves

                Biden makes serious AI regulation moves

                                            Francis Scialabba

President Biden issued an executive order on AI yesterday that is as comprehensive as a typical dad steakhouse order. The first-of-its-kind federal action aims to regulate an industry that has been mostly left to its own devices.

The biggest consequence? The developers of future advanced AI models will have to submit safety test results to Uncle Sam to prove they are not a public threat.


It’ll be a group effort

Different parts of the government will have a role in ensuring the tech doesn’t cause too much of a ruckus:

  • The National Institute of Standards and Technology will set benchmarks for safety testing, which the Department of Homeland Security and the Department of Energy will use to evaluate risks to national security and infrastructure.
  • The Department of Commerce will issue directions for watermarking AI-made content, so you can know if that Harry and Meghan divorce announcement is a deepfake.
  • The Department of Labor and the National Economic Council will study ways to mitigate the effects of AI on jobs.
  • The Department of Justice will get extra training and tech assistance to investigate cases of AI algorithmic bias. Landlords, federal benefits programs, and government contractors will receive guidance to ensure that they’re using AI equitably.

What else? The White House will develop guidelines for federal agencies to strengthen AI privacy protection practices, but Biden said it’s up to Congress to pass consumer privacy rules for the industry. The order also aims to standardize the government’s use of AI by streamlining the process for federal agencies to buy AI tools and hire experts. And it provides resources for small developers and researchers in an attempt to prevent a handful of corporate behemoths from dominating the industry.

But…the order is only a starting point that establishes important norms. It consists primarily of recommendations and contains some guidance that isn’t enforceable by the executive branch, making it unclear how effective it’ll be in practice. And, of course, it has little power over international AI developers.—SK


Previous Post Next Post