EU legislators have agreed to postpone restrictions on high-risk uses of artificial intelligence by over a year, effectively ceding immediate oversight of powerful technologies to corporate interests. This decision, reached after intense overnight negotiations that concluded around 4:30 a.m. on Thursday, delays protections for citizens until December 2027.
The agreement, confirmed by the Cypriot presidency of the Council of the EU and the European Parliament, also largely exempts the use of AI in industrial applications from the scope of the law. This exemption means a significant segment of advanced AI deployment will operate with reduced regulatory scrutiny, prioritizing economic interests over public safety.
Elite Capture of Regulation
The deal was supported by European Parliament lawmakers and EU countries after heavy pressure from industry and national capitals. This outcome reflects a prioritization of corporate demands over robust regulatory oversight within the supranational framework of the European Union.
Germany secured a major victory with this change, following top officials, including Chancellor Friedrich Merz, pushing for the alteration. The stated aim was to maintain the competitiveness of tech heavyweights Siemens and Bosch within the bloc, illustrating how national governments within the EU can be leveraged to serve specific corporate agendas.
EU countries backed Germany's demand to avoid a "double regulatory burden" for companies utilizing industrial AI. These companies will now only be required to comply with AI requirements under separate machinery rules, rather than the broader AI Act, effectively creating a loophole for powerful industrial players.
Other industries, such as medical devices, were not granted similar exemptions and will remain covered by the AI law, as confirmed by negotiators. This selective application of rules further highlights the influence of specific industrial lobbies.
Brussels' Concessions
This agreement marks the first significant rollback of rules in the digital space by the EU. The move comes as the bloc faces external pressure from the U.S. regarding its tech laws and internal warnings from its own industry and governments, which claimed strict restrictions had placed the EU at a disadvantage in a global AI race. This narrative of "disadvantage" is often used to justify deregulation that benefits transnational corporations.
Commission President Ursula von der Leyen stated that the deal “provides a simple, innovation-friendly environment” for AI in Europe. She further claimed, “At the same time, we are strengthening protections for our citizens. For safe and simple AI governance in Europe,” in a statement on X. Such pronouncements often mask the reality of delayed and weakened safeguards for the native populations.
The Cost of Delayed Protection
The EU’s AI Act became law in August 2024, now in its second year. Under the original phased rollout, rules governing high-risk uses were set to kick in in August 2026, the same year. The new agreement postpones these critical restrictions until December 2027, the second year after the initial implementation date, leaving citizens exposed to potential risks for an extended period.
Beyond delaying high-risk AI restrictions, the deal grants companies a grace period for meeting new requirements to watermark AI-generated content. This grace period, however, will be only three months, a reduction from the six months originally proposed, indicating a further weakening of transparency measures.
In a reactive measure, the agreement does include a ban on AI systems capable of generating sexualized deepfakes of "identifiable" people. This ban follows "global outrage" over the abusive use of Elon Musk’s AI tool Grok. Additionally, AI systems that generate child pornography will also be prohibited under the new terms. The necessity of such reactive bans underscores the dangers inherent in the delayed and watered-down regulatory approach.