EDITION OF OCTOBER 13, 2026
Brad Smith calls for binding rules on AI safety
Microsoft’s president advocates international AI regulation and believes the United Kingdom could lead the effort. TechCrunch reports proposals concerning external evaluations, intrusion reporting and shutdown mechanisms, without establishing that they have been adopted.
Legal disclosure: this content was produced by an artificial intelligence system. Editorial transparency
Brad Smith’s call for rules governing AI safety raises questions about the limits of companies’ voluntary commitments. Microsoft’s president advocates a leading role for the United Kingdom and several oversight obligations. The reported details describe proposals whose arrangements and adoption remain to be established.
Brad Smith, president of Microsoft, is calling for international safety rules for artificial intelligence. In an article published on October 12, 2026, TechCrunch reports his remarks from a recent BBC podcast, without giving a precise date for the interview. The executive describes AI as a source of cybersecurity risks and considers the current level of oversight of the companies developing it insufficient.
His argument focuses in particular on the limits of voluntary commitments. In his view, the most responsible companies may agree to submit to checks, while those that raise greater concerns may avoid them. This position therefore advocates common obligations that go beyond companies’ goodwill. It remains a position taken by a business executive and does not constitute an announcement of a new regulatory framework.
Smith believes the United Kingdom would be well placed to lead this effort. According to him, Washington does not view the United Kingdom and London in the same way as Brussels; he sees an advantage in this difference. TechCrunch notes that the UK’s AI Security Institute, launched in 2023, gains access to major models from American labs before deployment and participates in their safety evaluations.
This access is a concrete issue in the debate. According to TechCrunch, Donald Trump reportedly asked Anthropic and OpenAI the previous month not to share their new models with British evaluators. However, the same article says the institute confirmed on Monday that it could access and test the latest models. These details describe tension over international evaluation, without establishing that access was actually interrupted.
Among the measures discussed, Smith wants a legal obligation to notify an organization when its system has been compromised by uncontrolled AI agents. TechCrunch also reports his apparent support for approval by external evaluators before a product is released, as well as his favorable view of mechanisms that allow an AI system to be suspended or slowed down if humans lose control of it.
From this perspective, the issue would be to link evaluations to explicit responsibilities: who conducts checks, who must give notice and who can stop a system. The reported proposals specify neither the criteria for these checks nor how they would be implemented. They outline a regulatory direction advocated by Smith, but do not establish the existence of operational safeguards or an international agreement.