The European Union’s proposed AI Act has ignited a contentious debate among researchers, policy analysts, and developers over its potential impact on open source artificial intelligence. The legislation, first introduced in 2021 by the European Commission, seeks to promote “trustworthy AI” through a risk-based regulatory framework. It imposes requirements for risk management, data governance, technical documentation, transparency, accuracy, and cybersecurity. While the aim is to balance innovation with accountability, critics warn that the draft text could unintentionally stifle the very research that has driven much of AI’s recent progress.
A recent analysis from the nonpartisan Brookings Institution cautions that the AI Act, as currently written, would create legal liabilities for developers of general-purpose AI systems, even when those systems are released as open source. Alex Engler, the Brookings analyst behind the piece, wrote, “This could further concentrate power over the future of AI in large technology companies and prevent research that is critical to the public’s understanding of AI. In the end, the [E.U.’s] attempt to regulate open-source could create a convoluted set of requirements that endangers open-source AI contributors, likely without improving use of general-purpose AI.”
The legislation does contain exemptions for certain open source projects used exclusively for research and with safeguards against misuse. Yet, as Engler notes, preventing open source systems from entering commercial products is difficult. Stable Diffusion, an open source image generator, was released with a license restricting certain outputs, but quickly found use in creating explicit deepfakes of celebrities. Such examples illustrate the challenge of controlling downstream applications.
Oren Etzioni, founding CEO of the Allen Institute for AI, has voiced strong concerns about the draft’s burden on open source creators. “The road to regulation hell is paved with the EU’s good intentions,” he said. He argues that open source developers should not face the same compliance demands as commercial software producers, noting that a single student building an AI tool could be unable to meet regulatory obligations and thus be deterred from sharing their work. Etzioni advocates for targeting specific AI applications—such as autonomous vehicles or interactive bots—rather than broad technology categories, citing the rapid pace of AI development and the slow nature of legislative processes.
Not all voices in the field see the AI Act as problematic. Mike Cook, an AI researcher with the Knives and Paintbrushes collective, supports stronger regulation of open source AI. “The fearmongering about ‘stifling innovation’ comes mostly from people who want to do away with all regulation and have free rein, and that’s generally not a view I put much stock into,” Cook said. He believes setting standards can demonstrate global leadership and encourage other regions to follow suit.
The EU’s risk-based framework already prohibits certain uses, such as state-run social credit scoring, and applies restrictions to “high-risk” systems like those in law enforcement. However, as Lilian Edwards, a law professor at Newcastle School and advisor at the Ada Lovelace Institute, points out, the Act places responsibility on downstream deployers rather than initial developers. “[T]he way downstream deployers use [AI] and adapt it may be as significant as how it is originally built,” she writes, adding that the legislation does not adequately address the range of actors involved in AI’s lifecycle.
From the industry side, Hugging Face CEO Clément Delangue, counsel Carlos Muñoz Ferrandis, and policy expert Irene Solaiman welcome consumer protection measures but criticize the Act’s vagueness. They highlight uncertainty over whether rules apply to pre-trained models or only to end-user software. “This lack of clarity, coupled with the non-observance of ongoing community governance initiatives such as open and responsible AI licenses, might hinder upstream innovation at the very top of the AI value chain,” they said in a joint statement. The team warns that heavy burdens on openly released foundational models could suppress incremental innovation and dynamic competition in emerging AI markets.
Hugging Face promotes governance tools such as responsible AI licenses and model cards detailing intended use and operational mechanisms. “Open innovation and responsible innovation in the AI realm are not mutually exclusive ends, but rather complementary ones,” Delangue, Ferrandis, and Solaiman stated. They argue that regulatory efforts should focus on this intersection, a view gaining traction within parts of the AI community.
With multiple stakeholders and competing visions for AI’s future, the EU’s regulatory path remains complex. The eventual shape of the AI Act will depend on how policymakers reconcile the need for safeguards with the imperative to preserve open, collaborative development.