Members of the European Parliament are poised to expand the range of artificial intelligence systems subject to upcoming regulation. Hadrian Pouget writes that a transatlantic collision can still be avoided.
Hadrien Pouget is a Visiting Research Analyst for the Carnegie Endowment for International Peace’s Technology and International Affairs Program.
European parliamentarians are considering adding prohibited or high-risk applications to the first draft list of artificial intelligence (AI) law.
The primary purpose of the AI law is to require programmers working on high-risk applications to implement documentation, testing, and other safeguards.
Congress will use more than originally proposed, including systems that are “likely to influence democratic processes like elections,” and broader categories such as “general-purpose AI” that can be integrated into a variety of applications, such as OpenAI’s ChatGPT. is also poised to rope in many applications.
Big Tech is predictably resilient. The US government has expressed concern. A wider range could increase tensions across the Atlantic over technical regulations.
So far, Washington has remained relatively silent as Europe cracks down on technology. Through the Trade Technology Council, Washington and Brussels are focusing on cooperation and expressing a desire to harmonize regulatory approaches to AI.
This spirit of cooperation could be put to the test as AI law expands. US companies and regulators were already concerned about the AI law. We feared that AI law would try to tackle too many applications and become ineffective and burdensome.
US officials are sensitive because the EU’s recent Digital Services Act and Digital Markets Act were seen as unfairly targeting US businesses. In such an environment, expanding the scope of AI law risks losing the neutral appeal it has enjoyed to date.
Imminent conflict must and can be avoided. US critics need to recognize that AI law is not the often imagined “one size fits all” solution. The law applies a set of general requirements to high-risk AI systems, but the requirements can and should be adapted to different applications.
The “right level” of accuracy or robustness is customized for different high-risk situations. This allows for flexibility in enforcement and potential alignment with international standards. Also, the US and EU have already identified standards as an area of cooperation.
Recent AI law proposals also provide an opt-out mechanism for companies that do not believe their AI systems pose a risk.
US attitudes are shifting in favor of regulation. Senate Majority Leader Chuck Schumer has announced his ambition to put together an AI policy framework as the popularity of ChatGPT has fueled debate.
While details are scarce, Schumer’s statement suggests unifying requirements for AI systems according to four guardrails: “who,” “where,” “how,” and “protect.”
At the same time, the Department of Commerce began requesting comments on the development of AI audits and assessments, and the FTC began issuing sharp warnings about AI systems.
From this perspective, Washington and Brussels enjoy the opportunity to work together. The US and EU can partner to develop technical standards and other specific implementation details that underpin each other’s regulatory approach.
The EU, which is ahead of the US in some of these discussions, should invite and encourage this cooperation. Conversely, US lawmakers and federal agencies should take note of her EU approach when implementing their own requirements.
The European Parliament still has to approve the document and negotiations are protracted. EU lawmakers said he reached a political agreement on April 27, with a major committee vote scheduled for May 11.
A parliamentary vote will be held in mid-June. Even then, the parliamentary position must be coordinated with the Council of the European Union (which is made up of the relevant ministers of the Member States). Significant changes remain on the table.
The EU expansion (or threat of expansion) of AI law is understandably unnerving for Americans. The US should bring targeted proposals to the EU as the domestic conversation on AI risks matures.
A productive dialogue on AI regulation continues to be possible and necessary.
