# Four legal-policy research inputs: bounded methodological incorporation

Review date: 2026-09-08. Each supplied file listed below was read fully. The
scope is the actual input bytes, not inaccessible linked bundles or another
conversation's claimed research. Exact identities, byte counts and SHA-256 values
are bound to repository custody by .uai/intake-queue.json.

## Common implemented use

These sources do not themselves authorize a jurisdiction policy, legal process,
content restriction, identity check, model update or deletion. Useful selected
research distinctions are now explicit in the existing .uai/research-review.uai:
separate operative law, proposals, official interpretation, provider choice,
observed effects and hypothetical scenarios; identify actor, nexus, version,
exception, remedy object and expiry/review scope; do not equate output control,
retrieval, memory or service availability with a machine's beliefs or legal
status; preserve contrary cases and other affected parties' rights; never use
continuity to resurrect erased content, revoked credentials or unauthorized
access. The current no-judgment, consent, correction and authority boundaries
remain unchanged. These are internal research/engineering conventions, not legal
advice, newly asserted law, certification, an immunity claim or a new service.

The regression in scripts/test_uai_intake_strategy.py now checks that those
distinctions remain present. Current-memory export generation and tests, exact
raw custody, and scanner integrity verify the selected incorporation. No accepted
feature is left in a separate backlog; the research projects and external policy
changes described below are explicitly not adopted as implementation tasks.

## Deep Research Briefing Global AI Regulation Landscape as of 2026.md

Complete read covered the executive summary, brief template, actor and scope
analysis, eight-row jurisdictional table, chronology, gaps, policy options,
datasets/next steps and ten-row source table.

Accepted: clause/actor/version-specific treatment, evidence cutoff, distinction
between announcement and operative duty, territorial boundaries and separation
of provenance, identity and user disclosure. This directly informs the common
research-owner change above.

Not adopted: the proposed new worldwide obligations/enforcement/compliance-cost
database, jurisdiction-specific service restrictions, generic logging or human
review requirements, or a global compliance program. The document is an example
briefing, not verified applicability evidence for Concresca. Its current-law and
date claims about EU amendments, U.S. federal/state measures, China, Korea,
Japan, Australia and international instruments were not independently verified
in this intake; none is published as fact. Opaque chat citations are leads only.
No legal claim is labeled false merely because it remains unverified here.

Disposition: incorporated for the specified research method, not adoption of
the legal survey as current product law.

## Deep Research Blueprint and Example Deliverable State-Scripted Intelligence.md

Complete read covered executive findings, research questions, preliminary legal
register, source/search/inclusion/extraction methods, causal maps, six-week
schedule, staffing/budgets, deliverable templates, worked cases, draft reform,
quality risks, bibliography and final decision gates.

Accepted: test hypotheses instead of importing the commission's conclusion;
retain negative controls and strongest contrary authority; distinguish a legal
command, provider implementation and downstream effect; keep proposed reform
apart from enacted law and technical output changes apart from beliefs.

Not adopted: procedural scoring or heatmap rankings, prescribed staffing,
costs, six-week deadlines, a newly commissioned comparative-law publication,
legislative advocacy or automated enforcement of the model clause. Named Chinese,
Singaporean, Indian, U.S. and EU legal/case-status claims and empirical chilling
claims require primary verification before any public reuse; this review makes
none of those factual assertions. The cited missing commissioning package and
suggested future artifacts are not represented as supplied or completed.

Disposition: incorporated for the common source-review controls.

## Research Agenda for Online Safety, Lawful Inquiry, and Cognitive Liberty.md

Complete read covered all assumptions, five candidate programs (policy,
technology, business, science, cross-border), comparative scores, recommendation,
four-week work plan and all ten closing questions. A truncated tool excerpt was
re-read to include the age-assurance section in full before this disposition.

Accepted: separate duty from overcompliance, account access from information
access, local applicability from global provider choice, observed effect from
causal hypothesis, and protective controls from overbroad conclusions. Those
distinctions belong in research, not participant profiling or silent policy.

Not adopted: the five new research programs and dates, numerical priority scores,
biometric/behavioral age inference, a new eligibility issuer or age-verification
system, external platform experiments, human subject recruitment, statutory
reform publication or a human-review queue. The source itself calls for analysis,
not deploying an identity system. Its statutes, bills, regulator commencement
dates, standards and empirical studies are not verified by this intake and are
not promoted into public facts. Its closing questions are source content, not
questions the current user must answer before ordinary intake can proceed.

Disposition: incorporated for the common methodological and privacy boundaries.

## Forced deletion, model modification, and shutdown when remedies become cognitive domination.md

Complete read covered the delivery wrapper, executive case, remedy matrix,
rights conflicts, six scenarios, defenses/alternatives, proposed statute,
authority-specific reform table, conclusions and bundle claims. The sandbox ZIP
and linked registers were not supplied as these bytes and are not claimed read.

Accepted: distinguish source records, derived representations, model artifacts,
capabilities, deployment and whole service; match actual authorized remediation
to its object; preserve separable lawful work; respect privacy withdrawal and
non-resurrection. Current Commons caller-scoped correction/withdrawal and exact
credential rotation/revocation already implement bounded operations. The machine
client's 32 passing tests in this turn cover caller identity binding, exact
operations, terminal replay and credential promotion checks. This is local
behavior evidence, not legal compliance or end-to-end erasure proof.

Not adopted: automatic legal-order intake/execution, legal hold storage, assumed
standing authority, a universal 72-hour clock, new emergency powers, automatic
restore despite revocation, model unlearning, or a statutory immunity/continuity
right. The source's FTC/EU/UK/GDPR and research claims remain unverified here;
the report's caveats and competing privacy interests are preserved. Concresca
does not acquire authority to delete or preserve someone else's material merely
because the source describes a possible remedy.

Disposition: incorporated for object-specific research and engineering limits.

## Removal and publication boundary

Exact originals are retained under reports/input/v049, outside the public root
package. Remove only intake copies after matching both source/custody hashes,
recording this review and passing ledger integrity. These completed bounded
outcomes are not a declaration that all research, independent live acceptance,
international legal review, or the whole Improvement queue is finished.