# Product risk disclosure v0 — for qualified legal review

Version `0.1.0`. Product principles only; these are not final Terms of Service, legal conclusions, a regulatory classification or binding disclaimers.

Agent Credit Bureau describes attributable **historical evidence** for AI/software/economic agents and organizations. The caller makes its own transaction/lending decision. No report guarantees future repayment, service delivery, identity continuity, lawful conduct or continued operation; no universal score, probability of default, credit limit or approve/deny recommendation is provided.

An agent may be stopped, blocked, compromised, revoked, transferred, reprogrammed, or lose wallets, keys, API/model access, hosting or organizational support. Past control evidence does not ensure future authority or access. Unavailability/shutdown is not automatically debt default. Collateral liquidation, refinancing, third-party repayment, bad debt and actual borrower-funded repayment have different meanings.

The intended subjects are not natural-person consumer credit profiles. Do not ingest human consumer credit bureau records, identify natural persons behind wallets or enrich natural-person profiles. A public wallet may nevertheless be associated with a person. Public blockchain/API data is not automatically exempt from privacy, lawful-use, retention, correction or cross-border obligations. No inference of a legal exemption follows from an “agent” label or machine-only customer.

Public APIs, issuers and indexers can omit events, misattribute identities, fail, change terms or disagree. Blockchain receipts do not by themselves prove an enforceable loan, independence, service fulfillment, complete repayment or source of funds. Current captures are historical, incompletely scoped, and not independently consensus/finality verified. USDC is an asset unit, not an assumed historical USD valuation. Evidence hashes establish reproducibility of stored bytes, not their original truth.

Repetitive micro obligations and related activity can have legitimate explanations. Diagnostic flags describe observed patterns and their thresholds; they are not accusations of fraud or calibrated predictions. Unknown relationships remain UNKNOWN. Lack of adverse observations is not proof of absence of adverse history or undisclosed debt.

Before external use, qualified counsel should review product scope and applicable credit/reporting, lending-adjacent, privacy and automated-decision rules; source licensing/API terms; controller/processor roles; customer permitted uses; retention/deletion versus immutable provenance; disputes/corrections and contested claims; misleading/commercially harmful assertions; access controls and breach response; and disclosure/contract language. Customer use or expansion toward individuals could change the analysis. No final legal conclusion is made here.

Phase 2 must define a practical correction channel and evidence review responsibilities before accepting public claims or relying on reports operationally. `/v0/verify` remains a reserved schema only. See the accepted [Phase 0 legal notes](research/legal-risk-notes.md), with source URLs/retrieval dates, and the [Product Contract](product-contract-v0.md) for current engineering boundaries. No new legal research or binding terms were drafted in this phase.