A carton of milk is now subject to more regulation than an artificial intelligence system. Brad Smith, vice-chair and president of Microsoft, used this comparison in an interview published on 12 October by the BBC and reported by Reuters. “Every carton of milk you see in a shop is subject to safety regulations and inspections,” he said. “Today, your washing machine is subject to more safety regulations than AI.”
Two days later, at the International AI Summit in Dublin, the executive reiterated the point: “There are more safety regulations for a carton of milk than for AI. There are more safety regulations for your oven than for AI.” His assessment is straightforward: “The power of AI has advanced faster than its security. Security needs to catch up.”
The message comes from one of the senior executives at a company at the heart of the global race for artificial intelligence. Microsoft is one of OpenAI’s main partners and investors and is developing its own products based on Copilot and its Azure infrastructure.
Independent assessments
Brad Smith is not merely calling for additional voluntary commitments. He is advocating for independent assessments of the most advanced models, incident reporting requirements and the availability of an emergency mechanism should a system exhibit dangerous behaviour.
In Dublin, he also criticised the reluctance of some in the technology sector towards government intervention. “This aversion that some in the industry have to regulation is frankly something we need to get rid of,” he said, according to Reuters.
However, the debate remains constrained by a principle that Microsoft has been reiterating for several years: regulating without stifling technological development. In an interview with the Irish Times, Brad Smith argued that we must not create regulations that ‘stifle innovation’ or become excessively bureaucratic, whilst acknowledging the need for ‘a little more rule of law to ensure people’s safety’.
The other Brad Smith
For Luxembourg, this terminology relating to regulation and innovation brings to mind another Brad Smith. On 30 September 2004, the man who was then senior vice-president and general counsel at Microsoft arrived in Luxembourg for a two-day hearing before the Court of First Instance of the European Communities.
Microsoft was challenging the measures imposed a few months earlier by the European Commission in its antitrust case. Brussels had, in particular, accused the group of abusing Windows’ dominant position and had fined it €497m. Before the hearing, Brad Smith had made a statement which Microsoft still keeps on file. “We are very pleased to have the opportunity to be here today in Luxembourg,” he began.
The rest of the message was directed at the European Commission. According to him, the decision taken in Brussels would harm “hundreds of European companies, thousands of European software developers and millions of European consumers”. Above all, Brad Smith highlighted the risk that the regulator’s intervention posed to innovation. “We look forward to demonstrating that the path chosen by the European Commission will hinder innovation and slow economic growth – particularly in Europe,” he said on 30 September 2004.
Twenty-two years later, the same executive is calling on the authorities to catch up with a technology that, this time, has apparently outpaced the regulations designed to ensure its safety. The parallel has its limitations. In 2004, Microsoft contested the remedies imposed in a competition case. In 2026, Brad Smith is calling for security standards applicable to artificial intelligence. Defending innovation against certain antitrust obligations whilst calling for security standards does not, therefore, constitute a legal or economic contradiction. The change stems more from the role technology now plays in economies and societies.
In September 2007, Brad Smith was back in Luxembourg when the General Court essentially upheld the European Commission’s decision and the fine of €497m. In particular, Microsoft was required to provide its competitors with certain information to enable their products to be interoperable with Windows. The Microsoft case helped to shape the relationship between Brussels and the major US technology firms. It already centred on issues that continue to shape the current debate: to what extent can a regulator intervene in the design of a technological product, what specific obligations can be imposed on a dominant firm, and how can competition be safeguarded without stifling innovation.
Safety at risk
Artificial intelligence is now shifting some of these issues from competition to security. In his interview with Politico in Dublin, Brad Smith therefore advocates for an international framework capable of preventing a proliferation of incompatible regulations. He uses the example of commercial aviation, which is organised around mechanisms of mutual recognition and cooperation between states, enabling aircraft to cross borders.
He is, however, more cautious about the current European framework’s ability to accommodate all technological developments. When asked about the AI Act, he points out that legislation is drafted at a specific point in time, and is then confronted with technology that continues to evolve. It is then, he says, that we discover whether it ‘stands the test of time’ or whether amendments become necessary.
The issue goes beyond Microsoft. In Dublin, Henna Virkkunen, the European Commission’s Executive Vice-President responsible, amongst other things, for technological sovereignty, said that companies were not devoting enough time and resources to testing their models, given the intensity of global competition. Demis Hassabis, CEO of Google DeepMind, has for his part described several recent incidents involving AI systems as ‘early warning signs’. Brad Smith is now seeking to position Microsoft among those calling for additional regulations, whilst maintaining a consistent stance on their design: they must address risks without unnecessarily slowing down innovation.
In 2004 in Luxembourg, he warned that a European decision risked holding technology back. In 2026 in Dublin, he believes that technology has moved faster than the regulations designed to keep it safe. As for the milk carton, it already had its own regulations.
