EU AI Act Nears Completion, but Some Issues Are Still Open
Talks on EU AI legislation are expected to conclude by year's end, European Commission officials said at a Thursday briefing. The EC, European Parliament and European Council have held four "trialogue" meetings, including one on Tuesday, and everyone wants results quickly because of the unprecedented speed at which AI is developing, they said. International forums are also scurrying to develop guiding principles and codes of conduct, they said. Nevertheless, some issues remain controversial, such as biometric surveillance.
Sign up for a free preview to unlock the rest of this article
Timely, relevant coverage of court proceedings and agency rulings involving tariffs, classification, valuation, origin and antidumping and countervailing duties. Each day, Trade Law Daily subscribers receive a daily headline email, in-depth PDF edition and access to all relevant documents via our trade law source document library and website.
Europe's proposal treats AI as a product that should be safe and respect fundamental human rights, officials said. It identifies the level of risk in each application and targets regulation at the riskiest. High-risk applications are those with the greatest impact on people's lives, including such things as job recruitment.
Some issues remain unresolved, officials said. Among these are which AI applications should be deemed high risk, whether parts of the EC proposal should be fine-tuned, and what to do about emerging AI applications. Asked whether the talks could be derailed, one official said that while the political will is there to finish the legislation by the end of the year, nothing is agreed until everything is agreed.
The EU has created an AI Pact to encourage companies to engage with AI matters before the act becomes law, officials noted. Businesses could make voluntary commitments to anticipate the provisions of the AI Act, test best practices, and so forth. The pact should enable early adoption of the measure and save participating firms from having to make changes once the measure is in force, they said.
The EU is also engaged internationally at the bilateral and multilateral levels, officials said. Europe wants to support a global level playing field and promote good AI governance. It's working with the Group of Seven, G20, Council of Europe, Organisation for Economic Co-operation and Development, and Partnership on AI. G7 work is expected to produce by year's end guiding principles and a voluntary code of conduct aimed at AI developers; still under discussion is whether there should be a monitoring mechanism. The EU and U.S. have developed an AI road map.
Consumer groups said they're "alarmed" that the EU institutions may be about to adopt a "weak and unclear approach" to regulating generative AI systems such as ChatGPT. Citing leaked documents, the European Consumer Organisation said the suggested approach to dealing with generative AI would "trigger high uncertainty for regulators, consumers and the companies which might fall under the scope of the law."
"It is paramount that the use of AI systems that pose unacceptable risks to individuals and their fundamental rights are prohibited," the European Data Protection Supervisor said in a Tuesday "own-initiative" opinion on the AI Act. That includes barring their use for automated recognition of human features and other behavioral signals in public spaces, and categorizing people based on their biometric features.
The European Parliament wants a total ban on biometric surveillance; the EC wants to allow it in some circumstances such as countering terrorism; and some governments want it to be available for law enforcement within certain limits, the EC officials said. This is a key issue that must be settled during negotiations, they said.
The U.K. is going its own way. It will hold a global AI safety summit Wednesday to Thursday at Bletchley Park that will focus on certain kinds of AI systems based on the risks they might pose. Prime Minister Rishi Sunak announced Thursday that the U.K. will create an AI safety institute along the lines of the Intergovernmental Panel on Climate Change.