Smith’s remarks arrived as tech executives, European Union officials, and startup founders gathered to address the risks posed by autonomous agents. He specifically advocated for mandatory “safety breaks”—a mechanism allowing users to deactivate AI controlling critical infrastructure or household appliances. This proposal addresses emerging threats, including potential software “worms” capable of hijacking connected devices.

Beyond basic kill switches, Microsoft’s president highlighted the necessity for incident reporting requirements. If an AI agent gains unauthorized access to a home or business, the operator should face a legal obligation to disclose the breach. He further emphasized that while the industry must improve its internal testing and evaluation processes, self-regulation remains insufficient.

While pushing for oversight, Smith cautioned against overly bureaucratic frameworks. He noted that the European Union’s AI Act, spanning over 200 pages, risks stifling the very innovation it seeks to govern. Instead, he proposed a middle ground that balances necessary guardrails with the flexibility required for European firms to maintain their competitive edge in sectors like manufacturing and pharmaceuticals. Europe, he asserted, cannot afford to lag behind as AI reshapes every major global industry.