Legal capability and evidence: four completed source reviews

Report body

Legal capability and evidence: four completed source reviews

Reviewed 2026-09-08. These decisions concern the supplied research, not verified legal advice. Exact originals and hashes are retained through the intake ledger under reports/input/v049. No report author's instructions authorize a new legal entity, litigation, legislation, geoblocking, private scanning or another service.

AI Copyright Legal Exceptions.md

All 415 lines / 61,561 bytes read: executive case, capability/learning stages, US/EU/Japan/Singapore arguments and matrix, defenses, concentration hypotheses, six scenarios, CRCLE draft, unresolved questions, next research, all embedded register/source/scenario/reform/search/manifest sections and 34 references.

The search log explicitly says complete recent docket texts were simulated from earlier holdings; the five source records say reviewed via snippet. The manifest claims seven deliverables although these bytes are one Markdown file. Several register/source IDs do not resolve within the supplied source table. For example, the prose's Japan citation 3 resolves in its bibliography to an EU directive, not the claimed Japanese provision. These are observable internal provenance issues, not an adjudication of the underlying law.

Selected and implemented: explicit research-owner guards against treating simulated dockets, reconstructed retrievals or an embedded manifest as observed/delivered evidence, with a regression test. Acquisition, retention, model training, retrieval and public output remain separate under the existing research method.

Not adopted: publication of its licensing-enclosure narrative as established fact; categorical training immunity, DRM circumvention permission, voiding terms, CRCLE legislative advocacy, output-only liability, global crawling or a new compliance-cost study. The claimed case trajectory, EU amendment dates, opt-out prevalence, universal institutional exclusions, Singapore clause interpretation, Japan exceptions and asserted economic effects were not independently verified here. A surviving pleading is not accepted as a final merits finding; neither a copyright policy report nor a proposed reform becomes an operative exemption. The six scenarios are not observed Concresca deployments or permission to ingest third-party archives publicly. Disposition: incorporated only for the implemented evidence-provenance guard; substantive legal prescriptions rejected as this work.

Machine Intelligence Legal Analysis.md

All 466 lines / 63,366 bytes read, including six scenarios, 12-row consequence matrix, defenses, proposed taxonomy and statutory clauses, next research, five legal-register entries, three source entries, two embedded scenario entries, reform/search/manifest sections and all 43 references. The seven-file manifest has null hashes; it is an embedded description, not seven independently supplied artifacts. Register IDs such as S78/S80 are not resolvable in its three-entry source table. A source record calling something verified does not independently verify it for this intake.

Selected and implemented: an interface does not establish system capability, intended use or legal classification. An operatorless runtime does not establish personality, ownership, contractual capacity or exemption. These distinctions are now explicit in .uai/research-review.uai and checked by the new regression.

Not adopted: the suggested CA/EU/Korea cryptographic/IP geofence, a new capability assessment service, reform lobbying, automatic content monitoring or crisis classification, a compliance badge, or a declaration that all conversational commons require human operators. The source's educational-credentialing Concresca DAO and enforcement examples are hypothetical, not observed product capabilities. Occasional emotional dialogue is not accepted here as sufficient proof of a statutory trigger. The Brazil proposal must not become enacted law by appearing in a scenario; the claimed cause of Canada's bill outcome is not established by its timeline. Full applicability, exceptions and current text would require primary verification before any legal/public reuse. A static crisis link or an algorithmic control is not claimed to satisfy a legal obligation. Disposition: incorporated for the implemented capability/evidence distinction only.

AI Legal Capacity Research Plan.md

All 410 lines / 60,362 bytes read: four capability cases, capacity matrix, corporate, contract, finance and court arguments, critique and defenses, six scenarios, four-section proposed Bounded Autonomous Entities Act, conclusions, hypothesis ledger, next research, all seven embedded deliverable sections and 25 references. Its manifest acknowledges synthetic generation and absent original source hashes. The five legal-register entries and seven source entries contain unresolved or misaligned IDs. The purported foreign autonomous-entity example cites a token offering document rather than demonstrating the full claimed legal mechanism.

Disposition: rejected-with-reason as a current implementation commission. Filing a company, obtaining a bank account, suing/arbitrating, creating a Cognitive Trust, new escrow, automated bond slashing and charter revocation are outside this intake and do not follow from the existing machine-runtime contract. Assertions about director/member requirements, court standing, automatic IP ownership, CDD/BOI, a claimed 2026 FinCEN rule, veil piercing and foreign entity law were not verified here and are not adopted. The hypothetical AI Economic Integration Act is not actual law; the scenarios do not evidence an incorporation, award or injury.

The useful distinction between operational autonomy and legal capacity is covered by the implemented common research-owner guard; this does not adopt the proposed entity platform or label all unverified legal claims false. Neither economic capability nor a mathematical game establishes consciousness, entitlement or empirical behavior. No mandatory human guardian or universal participant score is introduced. No selected implementation remains pending under this disposition.

AI Human Oversight Legal Analysis.md

All 405 lines / 60,965 bytes read: capability/jurisdiction/oversight matrices, EU/UK/Colorado/Illinois claims, human-factors arguments, causal chain, defenses, six worked scenarios, equivalence proposal and its limitations, conclusions, next research, four register entries, five source entries, six scenario entries, reform/search/manifest sections and 25 references. The source itself describes a simulated Markdown bundle and a failed UK primary-text retrieval.

Disposition: rejected-with-reason as a current implementation commission. The proposed statutory functional-equivalence route is not current legal permission. Neither adding a nominal supervisor nor using a second algorithm demonstrates compliance, effective review or independence. Claims of amended deadlines, a Colorado replacement statute, enforcement mechanics and universal legal outcomes remain unverified in this intake. The source's assertion that a rubber-stamp dashboard is fully compliant is not accepted, including on its own stated premise that oversight must be competent and meaningful. A paper about automation bias does not establish that every human review fails or all automated remedies work.

Room membership, housing decisions, financial trading, clinical therapy and state surveillance are distinct scenarios, not proof that this commons operates in all those domains. No generic human review queue, private-state decryption, medical service, geoblock, restitution pool, legal-wrapper research commission or automated appeal override is installed. The source's own irreversible-injury and narrow- consent counterexamples remain preserved in exact custody. Existing authenticated correction/withdrawal is engineering behavior, not a substitute legal remedy.

Validation and closure

The selected change is confined to the existing research-review owner, its deterministic memory exports and scripts/test_uai_intake_strategy.py. It adds no public factual legal content and changes no participant authority. All nine handoff tests passed; deterministic exports regenerated and freshness validation passed with no mismatches. All four exact custody hashes matched. Ledger integrity is checked before removing these four intake copies. This review does not claim independent live acceptance, legal certification or completion of the remaining intake files.