Microsoft and Google executives are urging the European Union to reconsider whether its landmark AI law remains adequate, arguing that rapid advances in AI agents and growing concerns about the technology's risks have outpaced the two-year-old regulation. The comments, made on the sidelines of an international AI summit in Dublin, mark a notable reversal from an industry that has historically resisted regulatory oversight.
Microsoft President Brad Smith said European officials should revisit the relevance of the current rules as the technology continues to evolve.
"One of the things that is true of every law, of every regulation over time is it's written, it's interpreted and then the facts change, the technology evolves, and then regulations are often reinterpreted," Smith said.Demis Hassabis, co-founder of Google's DeepMind, made a similar point on stage at the summit, warning that legislation written years ago risks addressing problems that no longer reflect the real issues at hand.
The EU's Artificial Intelligence Act, the world's first comprehensive AI law, was largely negotiated in 2023 before generative AI tools and autonomous AI agents became widespread. Because the law classifies obligations by risk level rather than by specific technology, EU officials have argued it was designed to remain adaptable as AI capabilities advance. Agents — AI systems that can independently carry out multi-step tasks — are mentioned only briefly in the text, despite now powering more than half of AI workflows in Europe, according to data released by OpenAI this week.
European Commission officials have pushed back on claims that the law needs an overhaul. EU technology chief Henna Virkkunen said the framework is well-equipped to handle the risks posed by AI agents. EU Justice Commissioner Michael McGrath told reporters the bloc remains confident in its current approach, while acknowledging the importance of keeping rules adaptable as the technology develops.
Not every EU government shares that confidence. Hungary has told fellow member states that while the AI Act provides a solid foundation, it contains gaps that need addressing, according to an October paper. Among the concerns raised is that the law's obligations largely apply only once an AI model is placed on the market, potentially leaving unreleased or internally tested systems outside its scope. The paper also raised questions about whether regulators have sufficient visibility into how models are trained and tested, and about accountability when autonomous agents cause harm, suggesting these issues may require new legislation at the EU level.
The Czech Republic's AI envoy, Lukáš Kačena, described the law as a strong starting point that will inevitably require updates given how quickly the technology is changing.
"The AI Act will never be enough... This is never going to end," Kačena said.
This would not be the first adjustment to the law. Earlier this year, EU lawmakers agreed to delay implementation of certain provisions while adding new prohibitions, including a ban on sexualized deepfakes. That revision followed criticism that Brussels had moved too hastily in drafting the rules, having begun deliberations before ChatGPT's public launch in late 2022. The law's risk-based structure was intended to avoid exactly the kind of obsolescence now being raised by industry leaders, but the debate in Dublin suggests the question of whether Europe moved too fast, too slow, or simply not precisely enough remains unresolved.
Sources: POLITICO.eu