Brad Smith: Milk and Washing Machines Get More Oversight Than AI Companies

Microsoft President Brad Smith said that milk and washing machines are currently monitored more strictly than AI companies, and called for rules for the industry to be introduced urgently. He named the UK as the best place to develop them. The comparison is apt: household appliances require mandatory certification, milk is subject to sanitary control, while models that write code, answer medical questions and plug into business processes reach the market almost without external review.

One caveat up front: the source material only retells the gist of the speech, and we do not have the full text. Below is what was said, along with our analysis of the context, which helps explain why Smith is saying this now, and why about Britain in particular.

What exactly was said

Smith made two points. First, AI regulation is needed urgently, not at some point after the next incident. Second, the UK is better positioned than other countries to act as the hub where the rules take shape. The retelling does not name specific mechanisms, deadlines or a list of requirements, so it is too early to judge the details.

The speaker matters. Microsoft is among the largest AI developers and providers: the company builds models into Copilot, Azure and enterprise products. When a player like that asks for rules, it is signaling the terms on which it is willing to play.

Why London

Britain has a set of arguments going for it, and Smith appears to be leaning on them.

• In November 2023, Bletchley Park hosted the first international AI safety summit.
• The UK's AI safety institute, created in 2023, was renamed the AI Security Institute in February 2025. It tests frontier models before they are released.
• The country still relies on existing sector regulators rather than a single comprehensive law. That is faster, but it leaves gaps.
• Microsoft is investing in British infrastructure: in September 2025, the company announced investments of about $30 billion in the country through 2028, including AI computing capacity.

The last point is not necessarily a motive, but the overlap of interests is obvious. A country where a company has made large capital investments wants stable, predictable rules.

What the regulatory map looks like

Smith is speaking against the backdrop of three different models:

• EU: the AI Act, which has been phasing in since 2024. It takes a risk-based approach, with fines for prohibited practices of up to €35 million or 7% of global turnover.
• US: there is no single federal AI law; the rules are a patchwork of executive orders, state decisions and industry norms.
• UK: a flexible approach through sector regulators and a model-testing institute.

Smith's position is closer to the idea of mandatory standards for the most powerful systems. As early as 2023, Microsoft published its own AI governance blueprint, which included licensing for frontier models and safety requirements. The current statement continues the same line, only now with an address.

Why Microsoft wants this

Here is the sober assessment. Clear rules give large enterprise customers confidence that buying AI is safe. At the same time, they raise the barrier to entry: the costs of auditing, testing and documentation are easier for a company with a huge budget to bear than for a five-person startup. That does not mean Smith is insincere, but critics of regulation championed by big players already describe such calls as an attempt to lock in an advantage.

What to do in practice

Companies deploying AI in Britain or Europe should start now: keep a register of the models and use cases in play, document data sources, and appoint someone responsible for risk assessment. If rules do materialize, those with this record-keeping already in place will come out ahead. Developers should follow publications from the UK's AI Security Institute: its testing methods are the most likely to form the basis of future requirements.