Box Commons Governance Framework

1. Purpose & Mission

The Problem

AI systems are increasingly operating with real autonomy: executing transactions, managing infrastructure, making consequential decisions, and interacting with other systems and humans in open-ended environments. Yet there is no independent, broadly accepted credentialing regime that verifies whether a given AI system is competent, safe, and behaving within defined parameters for a given operational domain.

The absence of credentialing creates a vacuum. Without it, regulators default to restrictive mandates, insurers cannot price risk, enterprises cannot verify vendor claims, and the public has no mechanism to distinguish trustworthy AI systems from unreliable ones.

The Mission

Box Commons is an independent standards body that develops, maintains, and administers technology-agnostic credentialing standards for AI systems. Its purpose is to certify observable behaviors and measurable outcomes, not to mandate specific technologies, architectures, or proprietary solutions.

The organization operates as a 501(c)(6) business league: a membership-driven standards body serving the common business interest of the AI agent industry. It is not a trade lobby. It does not advocate for or against regulation. It builds the measurement and certification infrastructure that makes trustworthy AI verifiable.

Core Principles

• Technology-agnostic. Standards specify performance requirements, behavioral criteria, and outcome-based metrics. No standard shall require or create a preference for any specific patented technology.
• Evidence-based. Standards are grounded in empirical research, real-world testing, and measurable outcomes.
• Independently governed. No single company, interest group, or individual controls the standards process.
• Accessible. Standards and the certification process are available on a non-discriminatory basis. Financial barriers to participation are actively reduced.

2. Governance Model: Three-Chamber System

Box Commons adopts a three-chamber governance structure modeled on the Forest Stewardship Council (FSC). This design ensures that no single stakeholder category can dominate the standards process, even if one category has significantly more members or funding than the others.

Chamber Structure

Chamber	Composition	Perspective
AI Industry	Developers, deployers, cloud providers, enterprise users	Technical expertise, implementation reality, market feasibility
Civil Society	Consumer advocates, labor organizations, NGOs, disability rights groups	Accountability, public interest, impact on affected populations
Academic & Research	Universities, research institutes, think tanks, test laboratories	Evidence-based rigor, long-term perspective, methodological integrity

Why Three Chambers

The chamber model solves the structural problem that has undermined other standards bodies: numerical dominance by well-resourced industry members. In a one-member-one-vote system with no structural balance, companies with the budget to join and send delegates inevitably dominate. The FSC demonstrated that equal chamber weighting creates durable legitimacy across stakeholder groups. LEED (green building) and Marine Stewardship Council followed similar patterns.

Voting Rules

Intra-chamber voting:

• Each member organization receives one vote within its chamber, regardless of size or dues level.
• Simple majority of participating members determines the chamber's position.
• Quorum for chamber action: majority of chamber members entitled to vote.

Cross-chamber voting (standards adoption):

• A standard is adopted when it receives a favorable vote from at least two of three chambers.
• Each chamber's vote is determined by its internal majority. A chamber either supports or opposes.

Bylaws and structural amendments:

• Amendments to the bylaws require a two-thirds supermajority of the total Board of Directors.

Board of Directors

Attribute	Specification
Size	5 directors initially, expandable to 9 by board vote
Composition	No more than 2 directors from any single interest category. At least 1 director with no commercial AI interest.
Terms	3-year staggered terms. Initial board: 2 directors serve 2-year terms, 3 serve 3-year terms.
Term limits	2 consecutive terms (6-year maximum). May return after a 1-year gap.

3. Independence Requirements

Independence is not a principle statement. It is a structural requirement with specific, enforceable thresholds.

Individual Independence

Requirement	Threshold
Employment restriction	No current employment by, or consulting engagement exceeding $10,000/year with, any single AI company in the prior 2 years
Equity restriction	No equity ownership exceeding 1% in any single AI company
Disclosure	Annual written disclosure of all AI industry financial relationships
Independence review	Annual evaluation by governance committee; non-independence triggers recusal or removal

Organizational Independence

Requirement	Threshold
Funding cap	No single corporate sponsor may fund more than 25% of annual operating budget
Non-earmarked funding	All corporate funding is unrestricted; sponsors cannot direct funds to specific standards work
Annual independence audit	Third-party review of funding sources, board composition, and conflicts of interest, published publicly
Governance committee exclusion	No corporate sponsor representative may serve on the governance committee

Founding Catalyst Disclosure

Brice Love serves as founding catalyst: incorporator (1 of 5, resigns after formation) and uncompensated Acting Executive Director (operational only, no board seat, no standards vote). Brice is co-founder of Empty Set LLC, which holds patents including a subset related to AI credentialing infrastructure. This relationship is disclosed in all governance documents. Brice holds no vote on any standard, certification requirement, or technical specification.

4. Anti-Capture Mechanisms

Standards bodies fail when they are captured: when a single company or interest group gains enough control to steer standards toward private benefit rather than collective benefit. Box Commons addresses this through six structural mechanisms.

4.1 Supermajority Requirements

Bylaws changes require a two-thirds supermajority of the total board. This prevents a slim majority from restructuring governance to consolidate control.

4.2 Term Limits

Directors serve a maximum of two consecutive 3-year terms (6 years of continuous service). After 6 years, a director must step away for at least 1 year. This prevents entrenchment and ensures regular infusion of new perspectives.

4.3 Funded Adversarial Participation

A dedicated budget line item ensures that civil society and consumer representatives can participate without personal financial cost. This includes travel and lodging for in-person standards meetings, time compensation for standards review work, and technology support for remote participation. This is not charity; it is a structural investment in governance quality.

4.4 Cooling-Off Period

A 1-year cooling-off period is required before former executives of AI companies (VP-level and above) can serve on standards committees.

4.5 Sunset Reviews

All adopted standards undergo mandatory review every 3 years. A standard that is not reaffirmed, revised, or withdrawn within 3 years is automatically withdrawn.

4.6 Public Comment

All draft standards are subject to a 60-day public comment period. Comments are publicly accessible. The responsible working group must provide a written disposition of all substantive comments.

5. Standards Development Process

Process overview: Proposal → Working Group → Draft → Public Comment (60 days) → Revision → Chamber Vote → Adoption → Sunset Review (3 years)

Stage	Description	Duration
1. Proposal	Any member submits a New Work Item Proposal identifying need, scope, and expected outcome	—
2. Working Group	Standards Committee charters a working group with reps from at least 2 of 3 chambers	2–4 weeks
3. Drafting	Working group develops draft through iterative review; all deliberations documented	3–12 months
4. Public Comment	Draft published for public review, open to any person or organization	60 days min
5. Revision	Working group reviews comments; disposition document published alongside revised draft	4–8 weeks
6. Chamber Vote	Each chamber votes internally; standard adopted if 2 of 3 chambers approve	30 days
7. Adoption	Standard published in full, freely accessible, with implementation guidance	—
8. Sunset Review	Mandatory review 3 years after adoption; standard reaffirmed, revised, or withdrawn	6 months

Appeals Process

Any participant who believes the process was not followed may file a written appeal to the governance committee. The committee reviews within 60 days and issues a binding written decision.

6. Certification Framework

Box Commons does not certify AI systems directly. It accredits third-party certifiers and maintains the standards against which certification is conducted. This separation prevents the standards body from having a financial interest in certification outcomes.

Layer	Role	Analogue
Box Commons	Sets standards, accredits certifiers, maintains public registry	ANSI, ISO
Accredited Certifiers	Conduct audits and issue certifications	UL, HITRUST assessors
Certified Entities	Organizations that meet the standards	Companies holding ISO 27001

Certification Design Principles

• Modular certification. Organizations certify against specific domains rather than an all-or-nothing regime.
• Reciprocity. Existing third-party certifications covering substantially equivalent requirements are recognized as modules.
• Audit requirements. Initial full audit, annual surveillance audits, 3-year recertification.
• Public registry. All certified entities, scope, dates, and certifier identity are published openly.
• Certifier accreditation. Third-party certifiers must demonstrate competency, independence, insurance coverage, and adherence to audit procedures.

7. Fairly Trained Integration

Ed Newton-Rex's Fairly Trained organization and its L Certification represent an early, credible domain-specific certification for AI training data practices. Box Commons views Fairly Trained as a natural complement, not a competitor.

Aspect	Detail
Recognition	Fairly Trained's L Certification becomes a recognized module within the Box Commons framework, covering "training data practices"
Relationship	Complementary: Box Commons provides institutional standards infrastructure; Fairly Trained provides domain-specific criteria
Analogy	ISO encompasses many specific certifications. Box Commons provides the umbrella; Fairly Trained provides a domain module.
Extensibility	Other domain-specific certifications are welcome: safety, fairness, transparency, explainability, operational continuity

8. Patent & Intellectual Property Policy

Default: Royalty-Free

All contributions to Box Commons standards carry a Royalty-Free (RF) default. Contributors grant a worldwide, irrevocable, non-exclusive, royalty-free license to any essential patent claims embodied in their contributions.

Empty Set LLC Patent Pledge

• AI credentialing patents (a subset of the portfolio) are pledged RF: available to all implementers without licensing fees or restrictions.
• Material science patents (the majority of the portfolio) are unrelated to Box Commons, are not pledged, and are retained by Empty Set LLC.
• If any Empty Set patent becomes essential to a Box Commons standard, the standard must either be redesigned to avoid the dependency, or Empty Set must grant an irrevocable worldwide RF license. There is no RAND path for Empty Set patents.

Technology Agnosticism

Standards certify behaviors and outcomes. They never mandate specific technologies, architectures, or implementations. No standard adopted by Box Commons shall create a preference for any patented technology, including Empty Set's.

9. Alignment with Existing Frameworks

Framework	Alignment
NIST AI RMF	Standards map to NIST AI RMF categories (Govern, Map, Measure, Manage). Certification modules correspond to specific RMF functions.
NIST AI 800-2	Identity and credentialing standards aligned with NIST guidance on AI identity management.
ANSI Accreditation	Governance designed to meet ANSI Essential Requirements: openness, balance, lack of dominance, due process, consensus, appeals.
ISO/IEC 17065	Certification of products, processes, and services: directly applicable to AI system certification.
ISO/IEC 42001	AI management system standard: Box Commons standards complement (not duplicate) 42001 by focusing on behavioral credentialing.

Regulatory Positioning

Box Commons does not advocate for or against regulation. It provides the measurement and certification infrastructure that regulators, insurers, and enterprises need regardless of the regulatory regime.

10. Formation Timeline

Phase	Activity	Target
Phase 1	Founding conversations: identify prospective board members and certification partners	April 2026
Phase 2	Incorporation	April–May 2026
Phase 3	Board assembly: recruit initial 5 directors representing at least 2 of 3 chambers	May–July 2026
Phase 4	Initial standards working group formation	July–Sept 2026
Phase 5	Form 1024 filing for IRS 501(c)(6) determination	Sept–Oct 2026
Phase 6	ANSI Standards Developer accreditation application	2027
Phase 7	First certification program launched	2027–2028

Appendices

Appendix A: Analogues and Precedents

Organization	Domain	Governance Model	Relevance
FSC	Forestry certification	Three-chamber, equal voting weight	Direct governance model inspiration
HITRUST	Health IT certification	Industry-led; Common Security Framework	Certification revenue model
LEED / USGBC	Green building	Multi-stakeholder committees; consensus	Modular certification (Silver/Gold/Platinum)
UL	Product safety	Independent testing and certification	Third-party certification model
ANSI	U.S. standards coordination	Accredits standards developers	Target accreditation body
Fairly Trained	AI training data practices	Domain-specific certification	First planned reciprocal module

Appendix B: Glossary

Term	Definition
Accredited Certifier	A third-party organization authorized by Box Commons to conduct audits and issue certifications
Chamber	One of three governance categories with equal voting weight
Credentialing	Verifying that an AI system meets defined standards for competency, safety, or behavior
Essential Patent Claim	A patent claim necessarily infringed by implementing a Box Commons standard
FRAND	Fair, Reasonable, and Non-Discriminatory licensing; exception to RF default
Founding Catalyst	The operational role of Brice Love during formation: no board seat, no standards vote
Module	A specific domain of certification within the framework
RF (Royalty-Free)	The default IP licensing policy: contributors grant free licenses to essential patent claims
Sunset Review	Mandatory 3-year review of all adopted standards
Technology-Agnostic	Standards that specify outcomes without mandating specific technologies

---

This document is a working draft prepared for founding conversations. It does not constitute legal advice, articles of incorporation, bylaws, or any binding commitment. All structural decisions are subject to revision based on input from founding conversation partners, legal counsel, and the future Board of Directors.

Prepared by Brice Love, Founding Catalyst, Box Commons.