The race to achieve human-level artificial intelligence is moving so fast that the creators themselves are officially asking for a referee. In a personal manifesto published on X titled “A Framework for Frontier AI and the Dawning of a New Age,” Google DeepMind CEO and co-founder Demis Hassabis called on the United States to spearhead a dedicated AI standards body.
The Nobel laureate warns that artificial general intelligence (AGI) is likely only a few short years away. Future systems could introduce massive cybersecurity vulnerabilities alongside severe biological and nuclear risks. So, Hassabis believes Washington must move away from improvised, case-by-case reactions and establish a permanent, systematic playbook before the end of the year.
The FINRA blueprint for tech
Hassabis does not advocate creating a slow, heavy-handed government bureaucracy. Instead, he suggests modeling the new agency after the Financial Industry Regulatory Authority (FINRA)—the private, industry-funded watchdog that polices Wall Street under federal oversight. Under this framework, the major tech labs would foot the bill, providing the substantial financial backing required to pull in world-class engineering talent. It would also fund the massive computing resources needed for deep testing.
The watchdog would feature a majority-independent board combining independent technical experts, government officials, and open-source representatives. As detailed in an interview with Axios, frontier labs would initially submit their advanced models to the body on a voluntary basis for a 30-day pre-release review window. During this testing phase, experts would check the software for dangerous autonomous capabilities, attempts to bypass digital safety guardrails, and signs of digital deception while ensuring images carry proper watermarks.
Once the regime proves robust, compliance would become mandatory for any foreign or domestic model launching in the US market.
Clashing views on who hits the brakes
This proposal follows a quiet, months-long briefing tour where Hassabis pitched the framework to the Trump administration, European officials, and competing tech firms. The push comes on the heels of chaotic software standoffs last month. Recall that the US administration abruptly froze Anthropic’s Mythos and Fable models under export controls, forcing weeks of ad-hoc negotiations. Seeking to avoid a similar fate, OpenAI recently chose to restrict its newly released GPT-5.6 model to government-vetted partners at launch.
While there is broad industry agreement that the status quo is broken, leaders diverge on how to fix it. Hassabis favors a collaborative public-private model, but Anthropic CEO Dario Amodei has advocated for a stricter, FAA-style agency with the direct legislative power to block unsafe software outright. Meanwhile, American lawmakers are paying close attention as competitive pressure mounts from highly efficient Chinese open-weight models like DeepSeek and Z.ai. These are rapidly gaining traction among cost-conscious domestic startups.
The post Google DeepMind Chief Proposes Wall Street-Style Watchdog to Police Frontier AI—Led by the US appeared first on Android Headlines.
