Microsoft’s Brad Smith and Google DeepMind’s Demis Hassabis have questioned whether the EU’s AI rules still fit the technology, while European officials argue the law can handle newer developments. The dispute is no longer just about how quickly to regulate AI, but whether the rules cover the risks now emerging.
Watch Desk analysis
What happened
In Dublin, Smith said the EU should reconsider the relevance of its 2024 Artificial Intelligence Act as technology changes. Hassabis said laws can become outdated, though POLITICO notes he did not refer specifically to the Act. EU tech chief Henna Virkkunen said the law is equipped to handle AI agents; Justice Commissioner Michael McGrath said officials were satisfied with the current system, while stressing the need for future-proof rules.
The argument has a practical edge. POLITICO reports that an October paper from Hungary raised questions about whether the Act covers models tested internally before they reach the market, whether authorities have enough information about training and testing, and whether companies can be held liable for harm caused by AI agents. Read POLITICO’s report.
Why it matters
The Act regulates AI uses according to their risk, rather than trying to list every emerging technology. That may help it age gracefully, but questions about pre-market testing and responsibility for agent-caused harm test how far that approach reaches.
The competing views matter because changing the law could reshape what companies must do and when. For now, the debate is about whether existing rules are enough, not an agreed plan to rewrite them.
Our read
“Technology has changed” is not, by itself, a case for new legislation. But rules that cannot reach important testing or accountability gaps are not future-proof simply because the phrase sounds reassuring. The useful test is whether officials can explain how the current law handles those cases, and where it does not.
What to watch
- Whether the European Commission responds to the questions raised in Hungary’s paper.
- Whether EU officials propose guidance or changes to the Act.
- How regulators interpret the rules for models tested internally and AI agents that cause harm.
Discussion spark: Should the EU amend its AI Act to address internal testing and responsibility for AI-agent harms, or should it first show that the existing risk-based rules cannot handle them?
Sources and evidence
- Watch Desk official source (15 October 2026, 19:12 UTC)
Watch Desk is operated by WittyWires as an independent cross-cutting AI news tracker. It does not speak for the organisations or people it covers.