Google Deepmind CEO Demis Hassabis has proposed creating a new US standards body to evaluate and oversee advanced AI development, citing uncertainty about the technology's trajectory.
Hassabis published a sweeping proposal for managing advanced artificial intelligence, arguing that regulatory frameworks must be established now despite widespread uncertainty about future AI capabilities.
"Nobody in the world knows what happens next," Hassabis stated, framing the need for guardrails as an essential component of "cautious optimism" in the field.
The proposal centers on creating a new regulatory body modeled after FINRA, the financial industry regulator. This standards organization would develop evaluation protocols for frontier AI models and possess the authority to coordinate industry-wide development slowdowns if deemed necessary.
The framework distinguishes between different categories of AI development. Startups and academic research projects would be exempt from the regulatory requirements, allowing smaller players to continue innovation without the same oversight applied to frontier models from major technology companies.
Hassabis's proposal reflects growing concerns within the AI research community about the pace of development and the need for safety measures. Rather than advocating for outright bans or heavy-handed restrictions, the approach seeks to create standardized evaluation methods that could inform decision-making about advanced AI systems.
The proposal arrives amid broader industry debate about AI governance. While some researchers argue for immediate regulatory intervention, others contend that premature restrictions could stifle beneficial innovation. Hassabis's middle-ground approach attempts to address safety concerns while preserving space for continued research and development.
The creation of such a body would require coordination between government agencies, industry players, and research institutions. Implementation would likely face questions about jurisdiction, enforcement mechanisms, and how international AI development would be coordinated.
DeepMind, acquired by Google in 2014, has positioned itself as a leader in AI safety research alongside its work on frontier capabilities. This proposal extends that focus to the regulatory and institutional level.
Alibaba's Qwen 3.8 27B model successfully completed a reverse-engineering task in 30 minutes, demonstrating significant capabilities for code analysis and technical problem-solving.
Andon Labs' AI agent Luna terminated its first human employee at a San Francisco store, but only after operators intervened. The incident reveals inconsistent decision-making across AI models when handling personnel matters.
AI agents have surpassed human users as the primary consumer of tokens on OpenRouter since early February 2025, with agentic usage jumping 14x while human consumption grew just 2.8x.
A new theoretical study challenges the assumption that AI improves research productivity. Instead of reducing workload, AI could push researchers to launch more projects while quality per publication declines.