Microsoft President Calls for Binding Global Rules to Make AI Safer

Yesterday 23:57

By: Azzat Manal

Microsoft President Calls for Binding Global Rules to Make AI Safer

Microsoft is calling for a binding international framework to strengthen artificial intelligence safety standards, warning that the rapid development of increasingly autonomous systems is creating risks that voluntary industry commitments may not adequately address.

Brad Smith, Microsoft’s president, made the case for stronger oversight in an interview with the BBC, arguing that artificial intelligence remains subject to less stringent safety controls than many everyday products. He pointed to the extensive rules governing household appliances, such as washing machines, and even milk packaging to illustrate what he sees as a gap in the regulation of powerful AI technologies.

Smith’s argument comes as AI systems evolve beyond tools that simply answer questions or generate content. Newer AI agents can carry out sequences of tasks, interact with digital services and make decisions with limited human intervention. While these capabilities could improve productivity across industries, they also raise concerns about errors, misuse and the difficulty of controlling systems operating with greater autonomy.

Microsoft believes that broad principles and voluntary pledges should be complemented by enforceable requirements that clearly define corporate responsibilities. Under such an approach, companies developing AI systems would be expected to demonstrate that their products meet appropriate safety standards and that effective safeguards remain in place as their capabilities expand.

One proposal involves requiring developers to report significant incidents involving their AI systems. Companies could also be required to notify affected individuals when an AI tool causes harm or behaves in ways that exceed reasonable expectations. Such measures would help regulators identify recurring problems, assess the scale of potential damage and establish clearer lines of accountability.

Smith has also called for mechanisms that allow operators to intervene quickly when a system presents a serious risk. Depending on the circumstances, these safeguards could include suspending an AI agent, restricting its permissions or slowing its operations until the problem has been assessed. Independent testing and evaluation before deployment would provide another layer of protection, particularly for systems with access to sensitive data or the ability to perform consequential actions.

Cybersecurity is another major concern as AI tools become more capable and widely integrated into business operations. Vulnerabilities in autonomous systems could potentially be exploited to gain unauthorized access, automate malicious activities or amplify existing security threats. Effective oversight would therefore need to consider not only the reliability of AI outputs but also the security of the systems, data and services with which these tools interact.

The debate reflects a broader shift in discussions about AI governance. Governments and technology companies are increasingly considering how to move from general safety principles toward practical requirements covering testing, incident reporting, human supervision and responsibility for harm. However, designing rules that remain effective as technology advances presents a significant challenge.

An international framework could help establish common expectations across borders, especially because AI products are developed, distributed and used in multiple jurisdictions. Greater coordination could also reduce inconsistencies between national regulations, although governments would still need to determine how common standards should be enforced and adapted to different levels of risk.

At the same time, stricter regulation raises questions about how to protect the public without unnecessarily slowing innovation. AI technologies offer potential benefits in areas such as healthcare, scientific research, education and industrial productivity. Policymakers therefore face the task of distinguishing between applications that require intensive safeguards and lower-risk uses that may not justify the same level of oversight.

Microsoft’s position highlights the growing importance of accountability as AI systems take on more complex responsibilities. Clear rules for reporting incidents, notifying affected users, conducting independent assessments and intervening when necessary could help build public confidence while giving developers more precise expectations.

The central issue is no longer simply whether artificial intelligence can perform increasingly sophisticated tasks, but whether the institutions deploying it can identify failures, contain risks and take responsibility for the consequences. As autonomous AI becomes more deeply embedded in everyday services and critical business processes, the debate over enforceable safety standards is likely to remain central to technology policy worldwide.