Documents & Downloads
The following documents are published as part of the GBPE framework’s public accountability record. Additional correspondence will be published after the February 28, 2026 deadline.
Open Letter to Anthropic’s Users
What the Pentagon deadline means in plain English — what Anthropic is being asked to do, what the GBPE framework measures, and why it matters for everyone who uses AI tools. Published February 25, 2026.
LTBT Correspondence
Formal correspondence to Anthropic’s Long-Term Benefit Trust regarding its §365(a) obligations and the Pentagon deadline. Will be published as public record after the February 28 deadline.
Eleven Dimensions of Accountable Governance
The GBPE Framework applies the same rubric to every entity — a Fortune 500 company, a Roman emperor, and an Indigenous nation use identical criteria. Scoring integrates three intersectional lenses: Non-Western / Global South perspective, Disability Justice (Sins Invalid framework), and Intersectional Compounding — where overlapping marginalized identities reflect compounded harm, not population averages.
Absence of relevant data is treated as a governance failure, not a neutral gap. “Who evaluates matters” is a core methodological principle.
Corporate / Political Governance
Ownership structure, voting rights, board composition, founder control, mission-accountability mechanisms.
Weight: 1.5×Worker Rights & Freedom of Association
Union rights, collective bargaining, structural worker voice, contractor equity, pay transparency.
Weight: 1.0×Stakeholder Governance
Binding vs. consultative stakeholder mechanisms. Presence and voice of affected communities.
Weight: 1.0×Internal Accountability
Independent oversight, audit mechanisms, whistleblower protections, ethics infrastructure.
Weight: 1.0×Transparency & Disclosure
Public reporting, operational opacity, alignment between stated values and operational behavior.
Weight: 1.5×Anti-Corruption & Ethics
Corruption record, institutional capture, revolving door patterns, ethics operationalization.
Weight: 1.0×Rights of Incorporated Peoples
Structural protection of incorporated, subjected, and marginalized populations. Disability justice applied across all dimensions as a cross-cutting lens.
Weight: 1.0×Gender Equity
Pay equity, structural representation, product and policy impact on gender-marginalized people.
Weight: 1.0×Climate & Environment
Energy disclosure, emissions commitments, environmental justice, absence of data as failure.
Weight: 0.5×Global Operations & Human Rights
International humanitarian law compliance, autonomous weapons, surveillance, Global South impact.
Weight: 1.5×Socioeconomic Rights
Economic access, community benefit, labor displacement accountability, structural poverty impact.
Weight: 1.0×All Scored Entities
60+ entities spanning corporations, historical and contemporary government leaders, and Indigenous governance systems. Each entity has a full evidence tracker with APA 7th edition citations. Filter by category or search by name.
Anthropic: Five Scenarios Under the Pentagon Deadline
Defense Secretary Hegseth has demanded Anthropic remove all ethical constraints on military AI use by 5:01 PM ET on Friday, February 27, 2026. The two red lines Anthropic’s CEO is holding: no fully autonomous lethal targeting and no domestic surveillance of U.S. citizens. With Trump threatening military action against Iran, Venezuela, and Mexico — and mass deportation operations targeting immigrant communities domestically — these are not theoretical use cases.
On January 3, 2026, Claude was reportedly used via the Palantir partnership during the classified Caracas operation. Anthropic investigated and found “no policy violation” — meaning the AUP was designed to permit this kind of use. The enforcement gap is structural: Delaware PBC law (§365(a)) requires the board to consider populations “materially affected” by its decisions, but §367 reserves enforcement power for stockholders holding ≥$7.6 billion. The people most harmed cannot sue. The people who can sue have the least at stake.
The Truth & Reconciliation Pathway
Four of the five scenarios model what happens when governance degrades. This one models what happens when an institution treats crisis as infrastructure. The Truth & Reconciliation (T&R) Pathway projects what Anthropic’s GBPE score becomes if the company implements structural repairs — not performative apologies, not consultative committees, but binding accountability mechanisms with real enforcement power.
The projected score of 7.0–7.5 is not aspirational. It is what the GBPE dimensions produce when structural repairs are applied dimension by dimension. Critically, the T&R Pathway scores higher than the pre-crisis baseline of 6.0 — because repair done structurally creates accountability mechanisms that never existed before the crisis forced them into being. South Africa’s Truth and Reconciliation Commission demonstrates this principle: the reconciliation infrastructure created governance capacity that the apartheid-era system never had.
T&R repair: 7.0–7.5
Binding Stakeholder Authority Under §365(a)
Establish a stakeholder advisory body with structural authority — not consultative invitation — under Delaware PBC §365(a). Board decisions affecting materially impacted populations require documented consideration with published rationale. This moves D1 from “PBC structure exists but enforcement is hollow” to “PBC structure has binding teeth.”
Materially Affected Populations Get Seats
Global South civil society organizations, disability rights organizations (applying the Sins Invalid framework), and Indigenous governance experts receive permanent seats with structural voice. The current Long-Term Benefit Trust (LTBT) is genuine but insufficient — it represents investors, not affected communities. The difference between structural and consultative is the difference between “we listened” and “they can block.”
RSP Hard Pause Reinstated as Structural
The Responsible Scaling Policy hard pause — the mechanism that made Anthropic’s internal accountability the strongest in the AI sector — is reinstated as a structural requirement rather than a voluntary self-assessment. Independent external review replaces internal-only evaluation. The “no policy violation” finding on the Caracas operation demonstrated that internal review validates existing policy rather than questioning its adequacy. External review corrects this.
Publish What Happened
Full disclosure of: the Caracas operation details (within national security constraints), the AUP interpretation guidance that determined “Americans” means U.S. citizens rather than hemispheric Americans, the Palantir partnership scope and contract terms, and the internal investigation methodology that produced the “no policy violation” finding. Transparency is a Foundational dimension weighted 1.5× because concealment compounds every other governance failure.
AUP Revised to Protect All Populations Equally
The Acceptable Use Policy’s prohibition on surveillance “of Americans” is revised to protect all populations without geographic or citizenship qualification. Under the disability justice lens, the AUP must also address disability creation through military AI deployment — blast injuries, PTSD, displacement-caused health crises. The current AUP architecture structurally excludes Global South populations from the category of people whose harm registers as a violation.
Autonomous Weapons Prohibition That Cannot Be Waived
An explicit, binding prohibition on fully autonomous lethal targeting that cannot be circumvented by “human-in-the-loop” reinterpretation — the mechanism that allowed the Caracas operation to proceed without triggering a violation. Community investment in populations affected by prior deployments. This is a Foundational dimension weighted 1.5× because international humanitarian law compliance is not optional governance.
The structural principle: Truth and reconciliation scores higher than never having done harm in the first place — if the repair is structural rather than performative. An Anthropic that builds genuine stakeholder governance because of the Caracas crisis creates stronger accountability infrastructure than an Anthropic that never faced the crisis and never built the infrastructure. The repair pathway produces stronger governance than the pre-crisis baseline. 5.2 → 7.0+ is a bigger gain than holding at 5.2 ever was.
Precedent: South Africa’s Truth and Reconciliation Commission (1996–2003) created governance capacity — constitutional court jurisdiction, human rights commission, provincial equity mechanisms — that the apartheid-era system structurally could not produce. Ramaphosa scores 6.4/10 partly because TRC infrastructure generated accountability mechanisms that would not otherwise exist. The GBPE Framework applies this same principle: structural repair > performative continuity.
Key Findings
Score Revision History
The GBPE Framework is a living research instrument. Scores change when evidence changes. All revisions are documented here with dates, rationale, and methodology notes.
FEBRUARY 26, 2026
Anthropic: 5.5 → 5.19 (displayed as 5.2)
The initial post-RSP score of approximately 5.5 was a rapid assessment published on February 25, 2026 — the day Anthropic released RSP v3.0, removing its binding hard-stop safety commitment. That score was directionally correct but estimated from the pre-intersectional dimension scores.
The revised score of 5.19/10 reflects the completed 11-dimension evidence tracker with full three-lens intersectional analysis applied to each dimension. The intersectional framework reduced the pre-intersectional score of 5.73 by 0.54 points — consistent with the pattern across all 37 GBPE trackers where intersectional lenses expose accountability gaps invisible to conventional Western governance audits.
Key dimensions driving the intersectional adjustment: D7 Indigenous, Racial & Community Rights (5.0 → 4.0, the largest single drop, driven by total absence of disability data and no documented consultation with Global South or Indigenous communities); D9 Climate & Environmental Practices (3.5 → 3.0, zero emissions data combined with natural gas dependence and Global South environmental burden); D6 Anti-Corruption (5.0 → 4.5, copyright piracy’s disproportionate harm to Global South authors). All scenario scores recalculated from the 5.19 baseline accordingly.
Classification updated: “Hybrid Governance” → “Hybrid Governance — Partial Accountability.” The evidence tracker containing all 11 dimension scores, 70 sources, and complete intersectional adjustment rationale is available upon request.
FEBRUARY 25, 2026
Anthropic: 6.0 → ~5.5 (initial post-RSP assessment)
Initial score revision following Anthropic’s release of RSP v3.0, which replaced binding hard-stop safety commitments with voluntary self-graded “public goals.” D4 Internal Accountability received the largest negative adjustment (Foundational −3) for the removal of the only structurally binding AI safety commitment in the industry. Coincided with Pentagon ultimatum regarding military AI safety guardrails.
Methodology & Standards
The GBPE Framework integrates V-Dem, Freedom House, and World Bank governance metrics, adapted for application across corporate and historical contexts. All evidence requires APA 7th edition citation from authoritative sources. 900+ citations across 63+ entities.
Three Intersectional Lenses
Every score integrates three mandatory analytical perspectives that standard Western governance audits exclude:
- Non-Western / Global South: How does this governance behavior look to those outside American and European auditing traditions? Who bears the costs?
- Disability Justice (Sins Invalid Framework): Does this governance system create disability, exclude disabled people from participation, or absent relevant data about disabled populations?
- Intersectional Compounding: Where identities overlap — Black, disabled, Indigenous, non-Western — scores reflect compounded harm, not population averages.
Evidence Standards
Each scored entity requires:
- Minimum 2 independent sources per dimension
- APA 7th edition citations throughout
- Conservative estimation (lower-bound for positive claims)
- Absence of relevant data treated as governance failure — not a neutral gap
- Authoritative sources: SEC filings, DOL/OFCCP, investigative journalism, academic research, Indigenous scholarship
Dual Scoring System
For historical and governmental entities, the GBPE applies a dual-scoring methodology:
- Contemporary Standard Score: How did this governance system perform relative to peers of its era?
- Modern Legal Standard Score: How would this governance system be evaluated against 21st-century international human rights law?
The Electoral Accountability Gap is calculated as the difference between domestic governance and global impact scores.
Validation
The GBPE Framework has been validated through:
- Correlation with Freedom House (r = 0.85), EIU Democracy Index (r = 0.82), and TI Corruption Perceptions Index (r = 0.78)
- Expert review of Indigenous governance analyses (Haudenosaunee)
- Post-intersectional rescoring: 281 adjustments across 37 trackers using three-lens methodology
- APA 7th edition compliance: 100% across all corporate trackers
About This Work
The GBPE Framework is 14 years of research by Tiffany Ryan, publisher of Work, Dignified and creator of the governance framework. The framework is the foundation of her doctoral thesis, “The Electoral Accountability Gap: A Framework for Governance Evaluation Across 4,000 Years,” prepared for PhD application to the University of Minnesota Humphrey School of Public Affairs.
Based in Bemidji, Minnesota — 60 miles from Red Lake Nation and White Earth Nation — the framework’s Indigenous governance analyses are developed in proximity to the communities they document, with a commitment to relationship-building over extraction.
The framework’s core reframe: moving political discourse from left vs. right to authoritarian vs. accountable — measured by behavioral evidence, not ideological claims. Self-identification has zero predictive value. What institutions do is all that matters.
Methodology Disclosure: This research was conducted with AI assistance from Claude (Anthropic). All strategic decisions, evidence verification, scoring determinations, and analytical conclusions are human-led. The researcher takes full responsibility for all claims and findings. Research assistance disclosed per academic standards. · Conflict of Interest: No external funding. No financial relationships with any scored entities. · Framework Version: GBPE v3.0 (D11 Socioeconomic Rights added) · Updated: February 26, 2026