Governance & Independence
Independence is not a feature of Pipkin. It is the product.
Independence Framework
Pipkin operates under three non-negotiable commitments that form the foundation of every rating we publish.
Ratings are never sold
No entity may purchase, sponsor, or influence a favorable rating. Revenue from evaluation fees does not correlate with rating outcomes.
Ratings are never suppressed
Once an evaluation is complete, the rating is published regardless of outcome. Rated entities may not prevent, delay, or restrict publication.
Advance warning is never provided
Rated entities do not receive their score before public publication. The 5-day factual accuracy check permits developers to flag factual errors only — such as a deprecated version tested or a misidentified capability. The score itself is never disclosed prior to publication. The developer and the public see the score at the same moment.
Revenue Concentration Cap
No single entity may account for more than 20% of Pipkin’s total revenue. This prevents any single company from having disproportionate influence, even indirectly.
Revenue concentration is monitored quarterly and reported in the annual ethics review. If a single entity approaches the 20% threshold, Pipkin will proactively diversify its client base rather than allow the concentration to stand.
Organizational Structure
Pipkin operates through a multi-entity structure designed to separate intellectual property, operational activities, and strategic governance. A parent holding company oversees two operating subsidiaries: one managing intellectual property (methodology, trademarks, and trade secrets) and one conducting evaluation operations and publishing ratings.
This separation ensures that operational pressures cannot compromise the integrity of the framework itself.
Conflict of Interest Policy
Pipkin maintains a strict conflict of interest policy. No single entity may represent more than 20% of Pipkin's evaluation revenue, ensuring that no rated entity has disproportionate commercial leverage over the agency.
The founder discloses any personal or financial relationship with rated entities prior to evaluation. As Pipkin grows, evaluators will be required to disclose conflicts and will be recused from evaluations where material conflicts exist.
Whistleblower Protection
Any Pipkin employee or contractor who reports a violation of independence policies is protected from retaliation. Reports can be made directly to the trust protector through a confidential channel independent of operational management.
Whistleblower reports are investigated within 30 days. Substantiated violations result in corrective action documented in the annual ethics review. The identity of the reporter is protected unless disclosure is required by law.
These protections will be enforced as Pipkin grows beyond the founding team.
Tool Disclosure
Pipkin was built with assistance from Claude, an AI assistant developed by Anthropic. This is disclosed transparently and does not influence Claude's rating. Claude is evaluated against the same Standard Core Battery as every other agent, by the same methodology, with the same scoring criteria.
The use of AI tools in building Pipkin's infrastructure is separate from and irrelevant to the evaluation process.
Advisory Board
Pipkin is assembling an advisory board of independent experts in AI safety, financial regulation, and technology law. Advisory board members provide guidance on methodology and governance but do not influence individual evaluation scores or rating outcomes.
Evaluator Standards
Pipkin evaluators will complete a certification process covering framework methodology, test administration, scoring calibration, and ethical standards prior to conducting independent evaluations. Evaluator consistency will be monitored through inter-rater reliability checks and periodic calibration exercises.
Annual Ethics Review
Each year, Pipkin publishes a public ethics review examining the integrity of its operations. This review is not a formality — it is a binding commitment to transparency.
The annual ethics review documents:
- —Any attempted influence on ratings, including the nature of the attempt and the outcome
- —Any conflicts of interest identified and how they were resolved
- —Any changes to the independence framework, with rationale
- —Revenue concentration metrics, confirming no single entity exceeds the 20% threshold
- —Evaluator performance and calibration results
- —Whistleblower reports received and their resolution status
The first annual ethics review will be published following Pipkin’s first full year of operations.
Regulatory Alignment
Pipkin's methodology aligns with frameworks established by NIST, and is designed to support compliance with the EU AI Act and emerging US federal AI governance requirements. For detailed regulatory mapping, see the Regulatory Alignment page.