Companies will be forced to disclose the use of copyrighted material for AI-related projects after the EU Parliament reached a deal on the provisional text of the world’s first AI rulebook.
MEPs agreed on the draft of the EU’s AI Act on Thursday, April 27th. The Act is designed to set regulatory standards for the industry and iron out any ethical or commercial ambiguities before new AI technologies take hold.
Under the terms of the Act, companies will be mandated to programme human oversight into new systems and a risk classification system will have to be established for all new projects. Previous EU attempts to regulate AI have been frustrated by the release of ChatGPT and new generative technology which nullified previous drafts of the AI Act. Some experts have expressed fears that the EU could suffocate the embryonic AI industry with the Act and limit the type of research allowed within the bloc.
The Act is not expected to come into force until 2024. It will be subjected to a committee stage, and face its first plenary vote in mid-June. The delay in legislating for fast-moving technology means that there is speculation that national parliaments may take the initiative as the European Parliament’s legislative session comes to an end next year.
Human rights groups such as Amnesty International have expressed concerns that the AI Act would not ban racial pattern recognition. Left-wing MEPs added last-minute safeguards to prevent the use of AI technology in predictive policing or border control. The AI Act also inserts provisions for data to be anonymised and not to be disseminated among third parties.
Legislators were keen for AI projects to be environmentally compliant. All new ventures are obligated to maintain strict records on their carbon footprint and comply with environmental standards.
The EU hopes to get ahead of the global curve on AI by becoming the global regulatory leader for AI technology. Some experts believe Brussels is merely compensating for its own lack of innovation through overregulation.
The EU and the U.S. signed a regulatory agreement in February to allow for cooperation in the development of AI. The Brookings Institute, however, has warned of a potential transatlantic divergence due to Europe’s overbearing regulatory approach.
Attempting to regulate the unregulatable, the AI Act can be seen as a lethargic response by bureaucrats to dominate an industry that they, ironically, have caused to flee Europe due to overregulation.
Similar to the Digital Services Act and General Data Protection Regulation (GDPR), the AI Act attempts to imprint the EU’s value system on technologies not yet created and scarcely understood. Chaos potentially beckons, with Europe facing the prospect of being left behind as the world and the AI revolution move on.
