"A Framework for Frontier AI and the Dawning of a New Age" is an essay published July 14, 2026 by Demis Hassabis, co-founder and CEO of Google DeepMind and Isomorphic Labs. It proposes that the United States establish a frontier-AI Standards Body modelled on the Financial Industry Regulatory Authority (FINRA) — a federally overseen public-private self-regulatory organisation — to develop and run capability assessments for frontier models. It is the most detailed governance proposal published by a sitting frontier-lab CEO in 2026 and became the reference proposal in the US pre-release-vetting debate (AI Pre-Release Vetting).
Framing
Hassabis opens with the claim that artificial general intelligence, defined as "a system that exhibits all the cognitive capabilities the brain has," is "probably only a few short years away," and characterises the present as "standing in the foothills of the singularity." He argues AGI "cannot be compared to standard technological breakthroughs, not even ones as consequential as the internet or mobile," likening it instead to "the discovery of electricity or fire," and estimates its impact at "perhaps 10x of the Industrial Revolution at 10x the speed."
The risk case is stated in three tiers: cybersecurity challenges already observed in frontier models; nuclear and biological risks that "may soon emerge as capabilities continue to advance"; and, "on the horizon," the problem of maintaining control of "increasingly agentic, recursively self-improving systems."
The essay's stated reason for acting now is a mismatch between capability and understanding under competitive pressure. Hassabis writes that the field is "locked in an extremely intense, multilayered commercial and geopolitical race," that "advances on the frontier are outpacing our understanding of the technology," and that "nobody in the world knows for sure what is going to happen from here, and even the experts disagree." From that uncertainty he derives the policy posture: "when there is a large degree of uncertainty and the stakes are this high, proceeding with cautious optimism is the sensible and correct strategy."
The proposed Standards Body
The proposal has the following structure.
Form and funding. A new Standards Body "modelled on a federally overseen public-private partnership or self-regulatory organisation, much like the Financial Industry Regulatory Authority (FINRA)," with a board "that includes independent leading technical experts and open-source representatives." Funding "would need to be substantial and likely mostly come from industry," justified as necessary "to attract world-class technical talent and provide the necessary compute resources for large-scale testing."
Scope definition. A model qualifies as "Frontier-class" by meeting thresholds "on a set of benchmarks determined by the Standards Body and regularly updated." Organisations holding such models are designated "Frontier Labs" and are "encouraged to adopt best practices, such as publishing model cards with technical details, maintaining strong internal cybersecurity, vetting key personnel, and providing sufficient resourcing for safety and security research." Non-frontier models "say from startups or academia, would be exempt."
The voluntary-to-mandatory ratchet. "Initially, Frontier Labs would voluntarily share models with the Standards Body for review up to 30 days before release. Once the assessment protocol is shown to be effective and robust, formalisation could quickly follow, meaning that Frontier Models would be required to pass it to be deployed in the US market." Labs would also work with the body on "critical post-release vulnerabilities."
Coverage. The framework "could apply to Frontier-class models no matter their country of origin or whether they are open or closed" — the provision that extends it to Chinese and open-weight releases.
Assessment content. Evaluations would cover "cybersecurity, biological threats and other high-risk domains," with agentic tests looking for "attempts to bypass safety guardrails or signs of deception," and best practices including "digitally watermarking AI-generated images and generating human-readable output tokens to understand model reasoning."
Evaluation independence. Benchmarks would be refreshed "perhaps quarterly to start," with saturated benchmarks deprecated. Initially developed "in consultation with Frontier Labs," the body "should build up the technical capacity to create its own held-out tests independent of the Labs to prevent overfitting," and promote "an ecosystem of third-party auditors."
Escalation. The regime "could be ratcheted up if the seriousness of the situation demands, including coordinating a slowdown in development among the Frontier Labs if deemed necessary."
Hassabis presents the US effort as "a strong starting point for creating shared international standards on Frontier AI."
Open questions the essay leaves
The closing section states that even solved technical problems leave "further complex economic and philosophical questions": "what sorts of new economic models will be needed to help everyone thrive in a post-scarcity world? What values do we want to live by, what will meaning and purpose be, and how might even the human condition itself change?" Hassabis states these "cannot and should not be left to technologists alone."
Reception
Hassabis said he had briefed the Trump administration, fellow lab leaders, and European officials, and wanted the body operational "before year-end" (Source: axios.com).
Gary Marcus, who advocated FDA-style preflight testing in 2023 Senate testimony, welcomed the proposal the same day while cautioning that transparency and independence remain essential given US government interest in taking equity stakes in AI companies (Source: garymarcus.substack.com). Industry endorsements followed from Mustafa Suleyman, Satya Nadella, and Jack Dorsey by July 21; Dario Amodei said he favours an FAA-style federal agency instead (Source: axios.com).
Two critiques focused on enforcement and coverage. Zvi Mowshowitz argued on July 19 that "you need an SEC to your FINRA" and that a voluntary regime exempting internal deployment is inadequate (Source: thezvi.substack.com). Andy Hall's July 19 analysis identified open-weight releases such as Kimi K3, shipped without frontier-lab-style guardrails ahead of a public weights release, as the gap in any scheme built on voluntary lab submission (Source: freesystems.substack.com) — a gap the essay's country-of-origin-and-openness-agnostic clause addresses in principle but not in enforcement.
The standing administration counter-position predates the essay: Sriram Krishnan said in an interview published July 3, 2026 that "there will not be an FDA for AI" under Trump, ruling out a centralized frontier-model regulator (Source: ft.com).
The proposal nonetheless gained governmental traction. Hassabis's plan to lobby Washington became public July 16, and Bloomberg reported July 17 that the US government was considering creating a FINRA-like watchdog to vet top AI models — developed with Treasury Secretary Scott Bessent, reporting to the SEC, and under review by White House Chief of Staff Susie Wiles (Source: bloomberg.com; bloomberg.com).
Provenance
The essay was published as an X article at https://x.com/demishassabis/article/2076957440109625718, a host that scraping tools cannot retrieve and that has no Wayback capture. The text summarized here was retrieved on 2026-07-27 from the author's own Substack cross-post at https://demishassabis.substack.com/p/a-framework-for-frontier-ai-and-the-dawning-of-a-new-age, which carries the same title, byline, and July 14, 2026 date. Title, author, date, and distinctive passages were corroborated independently across TechCrunch (verbatim quotes), Axios (author interview), The Verge, and The Deep View, which put the length at roughly 1,500 words. Full verification record: Wiki/_meta/queue/gap-scan/proposed-sources/hassabis-frontier-ai-framework-2026.md.
Relationships
- supports: AI Pre-Release Vetting — the reference industry proposal in that debate.
- contradicts: Sriram Krishnan — the "no FDA for AI" administration position.
- related: Demis Hassabis, Google DeepMind, Gary Marcus.
- related: AI Safety Cases and Frameworks, Open-Weight Frontier Models, AGI Timelines.