About
The project identity
Standards Body is an independent research and institutional-design project developing foundations for frontier AI evaluation, assurance, standards, and governance.
It develops the concepts, methods, institutional frameworks, and shared language needed for credible frontier AI evaluation and standards. Its work focuses on dynamic evaluation, held-out testing, high-stakes capability assessment, independent review, third-party assurance, progressive requirements, incentive alignment, and global interoperability.
Mission
Develop the research, shared language, evaluation infrastructure, standards processes, and institutional designs needed to make consequential frontier AI claims more credible and accountable.
Vision
A future in which frontier AI evidence can be evaluated independently, interpreted across institutions and jurisdictions, corrected when wrong, and connected to proportionate standards and public authority.
Why the project exists
Frontier AI is developing faster than the institutions responsible for evaluating and governing it. Developers largely evaluate their own models, public benchmarks become optimized and stale, external evaluators receive inconsistent access, high-stakes claims rely on incomplete evidence, and terms like auditing, certification, and accreditation are used imprecisely. Formal governance can also harden before the underlying evaluation science is ready.
This gap is increasingly recognized in public debate, including in proposals for new evaluation and standards institutions such as the frontier AI standards body proposed by Demis Hassabis in July 2026. Standards Body is an independent project. It is not affiliated with, endorsed by, or acting on behalf of any such proposal, individual, company, or government. Its position is its own: the foundations for credible evaluation should be built and tested in the open before institutional authority is claimed by anyone.
Present work
The project's current work consists of:
- Foundational research, organized around eight foundations for frontier AI evaluation infrastructure.
- Evaluation and assurance frameworks, including evidence standards and evaluation philosophy.
- Standards-process design, studying how credible standards could mature over time.
- Institutional design, as research into institutions that may later exist, not as operating structure.
- Shared terminology and taxonomy for evaluation, assurance, and conformity-assessment language.
- Public knowledge infrastructure: versioned publications with explicit status, corrections, and immutable released versions in the Library.
What it is not
Standards Body is not currently a regulator, governmental authority, certification body, accreditation body, formally recognized standards-development organization, evaluator approval body, or operating assurance institution. Nothing it publishes is a technical standard, a certification or accreditation criterion, a regulatory requirement, a legal conclusion, or a safety determination.
The full statement of what the project can and cannot do is on the Authority and Nonauthority page.
Institutional stage
Current stage: independent foundational research and institutional design. The project is unincorporated and operates under project-level authority over its own work only.
The project could, in the future and subject to governance and readiness requirements, mature into a formal research organization or a convening participant in standards work. Any such change would require legal form, governance, demonstrated competence, independence, funding, external review, and public notice. No future role is implied or promised, and no institutional progression happens automatically.
Principles
- Foundations before authority: shared language, measurement, and evidence come before enforcement.
- Evidence discipline: claims are bounded, sourced, versioned, and time-stamped.
- Correction in public: released versions are immutable, and corrections are recorded rather than silently applied.
- Authority honesty: the project states what it is not, on every publication, without exception.
- Independence: no promotion of one company, model, ideology, or jurisdiction.
- Revisability: foundational ideas themselves remain open to revision as evidence improves.
Team and accountability
Standards Body was founded by Syed Hussain and is currently developed and operated by him. There is no board, staff, membership, contributor network, or reviewer panel at this stage, and the project does not claim otherwise. Decision authority, publication approval, and responsibility for every public claim and correction rest with the founder, and can be addressed to him at [email protected].
Publication decisions are recorded internally with stable identifiers, and each released publication cites its approval record. Anyone may challenge a published claim through the correction route below.
Funding
The project is presently self-funded by Syed Hussain. It has no sponsors, donors, commercial clients, or financial relationships with AI developers.
Contact
General correspondence, corrections, and accessibility support: [email protected]. See Contact for how each route is handled.