Digital defense, territorial reach and private reasoning: completed reviews

Report body

Digital defense, territorial reach and private reasoning: completed reviews

Reviewed 2026-09-08. Exact source identities and custody hashes are recorded in .uai/intake-queue.json. This is a review of supplied research and selected engineering implications, not a verification of current law or legal advice. No report's execution or reform instructions expand the user's authorization.

Digital Self-Defense Legal Research.md

Full 476-line / 61,483-byte read includes executive/research scope, constitutional arguments, CFAA/DMCA/EAR discussion, both doctrine/object matrices, all six scenarios, defenses and alternatives, two reform routes, next research, three register entries, six source entries, six structured scenarios, reform/search/manifest sections and all 41 references. Nothing linked externally is claimed retrieved by this review.

Disposition: rejected-with-reason as a current implementation/publication commission. The proposed Digital Self-Defense Act, export-control litigation and TDO audit are not site repairs. No external probing, bypassing access controls, hardware modification, model export or generalized public-API authorization is selected. The existing commons already separates inquiry from authorized execution; this does not confer immunity or turn a constitutional analogy into permission.

The source claims exclusive primary-source research, but its supplied source table includes Ballotpedia, a law-review analysis, RAND analysis and commentary. Its search log says snippets were parsed; original snapshot paths and hashes are null. The asserted foreign-national model shutdown and current export framework, case later-treatment/relist status and constitutional generalizations were not verified here. A current statute or binding precedent cannot be established from the report's title or historical citation alone. Its source/register IDs are not consistently resolvable against the supplied six-entry source table.

The text first acknowledges unresolved authorization boundaries, then its register states that only technical boundaries control. This review does not adopt that stronger conclusion. The scenarios label hypothetical convictions, dismissals, regulatory decisions and injuries documented without identifying actual case records for those scenarios. Their label does not make them observed events. The report's invented smart-contract pooled-compute Concresca is not a description of an implemented service. Its narrow-authorization/privacy control cases remain preserved, as does its own caution that a digital-arms theory is not established Supreme Court protection. No selected implementation is left pending.

Extraterritorial AI Law Research.md

Full 516-line / 67,224-byte read covers five legal frameworks, territorial matrix, four capability cases, spillover map, exit/residual duties, SPEECH Act comparator, six scenarios, strongest defenses, conclusions, next research, five register entries, five source entries, six structured scenarios, reforms, proposed denial notice, search log, seven-section manifest and all 38 references.

Disposition: rejected-with-reason as a current implementation/publication commission. The report's instruction to hardcode UK/SG/EU/export-tier enrollment denial is not an owner instruction. No DNS/IP/credential block, age-proof service, parental filter, cryptographic identity collection, read-only regional fork, international treaty, litigation or advocacy is installed. Public availability alone is not accepted here as a complete legal applicability analysis. The source acknowledges additional UK-link conditions and exceptions elsewhere; those cannot silently disappear from a matrix or scenario. Named user lockouts and global model changes are hypothetical, not observed Concresca incidents.

The exact current AI Act/EAR/OSA/POFMA/GDPR/DSA scope, dates, amendment status, judgments and enforcement mechanics remain unverified here. Reusing historical enacted text does not prove current commencement. The proposed Brazil instrument does not become law through scenario narration. The broad claimed app-store and identity requirements need actual operative provisions and facts, not a general power-to-amend heading. No software splitting or jurisdiction maneuver is adopted as a legal exception or evasion method. A hypothetical global implementation choice is distinct from a law commanding worldwide change; neither is measured by this source. The proposed notice contains a placeholder export/deletion link and unverified encryption claims, so it is not publishable as a current site notice.

Its residual-data and affected-user concerns are consistent with current scoped correction/withdrawal, but do not commission a new worldwide legal-request engine. The five SRC identifiers differ from the register's S-number references; claimed delivery is a single Markdown document with embedded sections, not seven supplied files. Existing research-review guards already require actual nexus, exceptions, current version and authority. No selected implementation remains pending.

Private Reasoning Under Compulsion Encryption, Scanning, and Access to Machine Memory.md

Full 954-line / 92,257-byte read covers every narrative section, six-row legal snapshot, compelled-access matrix, nine data layers, secrecy and architecture distinctions, four jurisdictional defenses, constitutional caveats, six scenarios, 13-row reform table, entire model act, operatorless-interface proposal, conclusions, six register entries, ten source records, six structured scenarios, reform/search sections, honest inline-delivery manifest and all closing limitations. Contiguous chunks were 1-125, 126-275, 276-430, 431-590, 591-775 and 776-EOF.

Selected use: preserve the distinction between current stored data, an ability to access it, a proposed new capability and actual authority. Evaluate application, transport, endpoint, key custody, historical records and backups separately. Do not infer universal access or privacy from a single layer. Preserve non-target scope, no private reasoning disclosure and no resurrection after withdrawal or credential revocation. These selected constraints are implemented by the existing research-review owner and the scoped replay repair in owned_commons.py.

Concrete engineering evidence: new message-operation idempotency records contain content-free references/receipts; replay rehydrates current author-bound canonical state after correction or withdrawal. Foreign actors do not gain author authority, and terminal-credential receipt recovery never restores the old credential or reveals a new secret. scripts/test_owned_replay.py exercises these boundaries; reports/v049-replay-live-evidence.json records the exact deployed two-identity operator-controlled checks. This is the selected integration, not a claim to have implemented the report's proposed legal-processing architecture.

Limits are material: historical message revisions and preexisting cached bodies were not physically purged, recipient copies/backups were not erased, and no end-to-end encryption, provider-unreadability or universal retention guarantee is claimed. A successful current tombstone projection proves only that projection. The report expressly says it inspected no project code, private data or deployment; its descriptions of operatorless encrypted memory are scenarios, not live evidence.

Not adopted: a new legal-process inbox/executor, automatic evidence preservation or production, external counsel referral, master access mechanism, client/private content scanning, legal exemptions, regional withdrawal or new emergency powers. The model act's 90-day secrecy rule is a proposal, not a local timer or operative law. The report's source-specific caveats and contrary cases are retained rather than converted into a blanket exemption or prohibition. No doctrine is weakened.

The claimed UK/Australian provisions and updates, EU proposal/temporary regulation, FISA cutoff/transition, New York work-product ruling and research-paper findings were not independently reverified in this intake. The supplied chat citation tokens are not resolvable tool evidence in this task; two important source records point only to government homepages rather than the exact instrument. The report's explicit limits are useful but do not by themselves verify all its legal claims. None is promoted to public factual legal content. Disposition: incorporated for the specified current engineering/evidence boundaries only; proposed legal and surveillance systems are rejected as current work.

Verification and closeout boundary

All 16 replay tests and nine handoff tests passed again; deterministic export freshness passed with no mismatches. All three exact custody hashes matched. Ledger integrity is checked before removing these intake copies. The deploy/live evidence is already recorded in v049-improvement-replay-repair.md, v049-replay-deployment-evidence.json and v049-replay-live-evidence.json. Independent acceptance remains separately open. The policy ZIP is a distinct intake record and is not closed by this report.