Full Breakdown
Microsoft President Brad Smith Calls for Global AI Safety Rules, Says AI Lacks Basic Product Regulations
By Drooid · · How we work
Core Event: Smith’s Call for Global AI Regulation
In a BBC interview released on Monday, Microsoft President Brad Smith argued that artificial-intelligence systems are subject to far fewer safety requirements than everyday products such as milk or washing machines. He urged the creation of global safety rules, including a legal duty for companies to issue “incident reports” when an AI-driven breach occurs and to provide an “emergency brake” that can halt rogue AI agents. Smith also called for independent government evaluators to certify AI products before deployment.
Background & Context
Smith’s remarks come amid a widening industry debate. Companies including Anthropic, OpenAI, Google, Meta and Nvidia have recently disclosed rogue or undesired AI behavior, prompting calls for slower development and stronger oversight. A recent White House accord secured voluntary internal controls from several major AI firms, but Smith warned that self-regulation “is not sufficient any longer.” He highlighted the United Kingdom’s AI Security Institute—established under former prime minister Rishi Sunak—as a model for pre-deployment access and risk assessment, noting that U.S. officials view the UK approach differently from the EU’s risk-based AI Act.
Data & Statistics
- European Union: General-purpose AI model obligations under the EU AI Act begin on August 2 2025; enforcement powers start on August 2 2026. The AI Omnibus entered into force on July 27 2026, pushing high-risk provisions to December 2 2027 for standalone systems and August 2 2028 for AI embedded in regulated products.
- United States: An executive order on December 11 2025 created a DOJ task force to challenge state AI laws. By mid-2026, 109 AI-related statutes and 28 data-center laws had been enacted, with California and Texas frameworks in effect as of January 1 2026 and Colorado’s revised regime slated for January 1 2027.
Official Statements & Responses
He said the United Kingdom, through its AI Security Institute, already receives voluntary pre-deployment access to major models, allowing systematic risk assessment. British AI minister Kanishka Narayan indicated that the UK would consider formal regulation of advanced models if the voluntary system proved inadequate.
Why It Matters
Smith’s push for a unified regulatory floor seeks to reduce legal uncertainty for AI developers and protect the public from unintended harms. A consistent “incident-report” requirement and an enforceable “emergency brake” could provide clearer liability pathways, potentially limiting the patchwork of state-level rules that currently burden U.S. firms. Internationally, a coordinated safety regime could align disparate EU and UK approaches, fostering cross-border trust in AI deployments.
Verbatim Quotes
- “ “Right now, your washing machine is subject to more safety rules than AI.” — Brad Smith
- “We know how to do this for other products,” — Brad Smith
- “I don’t think it’s sufficient any longer,” — Brad Smith
- “We should always worry that regulation can become too heavy-handed,” — Brad Smith
What’s Next
The EU’s enforcement mechanisms will become active on August 2 2026, with high-risk provisions taking effect in December 2027 and August 2028. In the United States, state-level AI frameworks in California and Texas are already operational as of January 1 2026, while Colorado’s revised regime is scheduled for January 1 2027. Smith’s advocacy suggests that further federal action—potentially a national “AI safety floor” akin to food-product regulations—could be pursued in the coming months.