Passage annotations

A lawful query is evidence of inquiry, not by itself a confession of belief, character, or harmful intent.

This is a QAEDA normative presumption. It can be displaced by additional context or conduct; the topic alone is insufficient.

Longitudinal query data becomes more sensitive when combined with identity and behavioral signals.

The linked claim is supported rather than absolute; sensitivity and inference quality vary by dataset and context.

Request interpretation can become persistent profiling when query context is joined to broader user history.

This note distinguishes immediate relevance inference from durable cross-request person models.

Consequential ranking, refusal, sanction, or risk classification should expose reasons and meaningful correction paths.

This is QAEDA’s governance inference from due-process and auditability research; it is not a claim that every ranking decision requires formal adjudication.

Studying a dangerous or controversial subject does not establish endorsement of it.

The presumption protects journalism, scholarship, prevention, criticism, fiction, and ordinary curiosity while allowing context-sensitive intervention when evidence changes.

Safety can often target what an interaction materially enables rather than treating an entire field of knowledge as forbidden.

The linked position rejects both unrestricted operational assistance and topic-wide suspicion.

Some safety outcomes can be pursued while retaining less or no centralized raw inquiry history.

The claim is architectural possibility, not a claim that every safety function can be decentralized or made zero-knowledge.

Services can sometimes verify a needed attribute without collecting a full identity dossier.

Applicability depends on the eligibility problem, threat model, and credential architecture.

The safety-poison-pill thesis survives only in narrowed form: function creep and chilling effects are risks, not proof that all safety controls are surveillance.

This annotation preserves the adversarial report’s correction instead of retaining the stronger rhetoric.

Exceptional powers can expand beyond their original purpose, but persistence and expansion are not inevitable.

Rollback, sunsets, deletion, and institutional reform are part of the historical record too.

Deliberate information seeking and manipulative push-based amplification are different governance problems.

QAEDA uses this distinction to protect developing autonomy while allowing stronger safeguards against exploitative recommender design.

Age-threshold checks need not always disclose a complete real-world identity to the relying service.

This does not imply one credential design works for every jurisdiction or risk level.

The QAEDA Cognitive Liberty Audit defines ten separate categories rather than one ideology score.

This is a verifiable description of QAEDA’s published experimental framework, not external certification.

Transparent reasons and meaningful recourse make consequential automated power more contestable.

The claim is procedural: transparency does not guarantee correctness, but opacity removes an important correction mechanism.

The shorter Smedley attribution and the longer two-clause wording have different evidentiary status.

The shorter form has a credible secondary attribution; the added sentence remains unverified as Smedley wording in the reviewed corpus.

“Rebel or Die” is not verified as a canonical movement-level slogan of post-2022 Russian anti-war resistance.

The negative finding means no verifiable movement-level provenance was located; it does not prove no individual ever used similar wording.

“Нет войне!” (“No to war!”) has a dense dated documentary trail from 24 February 2022 onward.

This is the stronger historical anchor for the public essay’s account of Russian anti-war language.

Russian resistance to the war and mobilization was heterogeneous rather than one unified movement.

The source set distinguishes protest, exit, legal avoidance, conscientious objection, family activism, refusal, and a smaller destructive fringe.

Removing the inserted token and joining “reve” + “lation” reconstructs “revelation” without changing a letter.

This establishes the string reconstruction, not the identity of the referent.

Search adjacency among unrelated “Orange Revelation” uses does not establish one transmitted cross-domain concept.

One source deliberately explores an interdisciplinary motif; the source-critical report shows why that metaphor should not be mistaken for historical transmission.

A visibility system can materially restrict discoverability without deleting the source artifact.

The source corpus supports the mechanism distinction; it does not prove that every ranking loss is deliberate censorship.

Private reception is upstream of public expression and can be chilled by identity-linked surveillance.

The principle is stronger than any single legal doctrine; targeted process and necessary identity-dependent transactions remain distinct cases.

Censorship drift is a recurring institutional risk, not an inevitable outcome of every safety rule.

The source set contains both warning mechanisms and legitimate restriction frameworks; QAEDA rejects inevitability rhetoric.

Independent redundant custody reduces single-point erasure risk for important records.

Redundancy improves resilience but does not remove authenticity, privacy, copyright, or safety obligations.

Financial intermediaries can materially constrain lawful speakers even when they do not remove the speech itself.

Commercial enforcement is not automatically equivalent to state censorship; actor, coercion, alternatives, and due process matter.

Translation, omission, and controlled vocabulary can create different informational worlds across languages.

Specific translation claims require source-text comparison; ordinary linguistic ambiguity should not be mislabeled as censorship.

Personalized systems can create user-specific visibility boundaries, but capability alone does not prove intentional personalized censorship.

Controlled comparative evidence is required before attributing motive or viewpoint discrimination to a particular provider.

Some eligibility checks can disclose an attribute without disclosing a full identity dossier.

This is an architectural possibility, not a claim that every jurisdiction or service can safely eliminate identity requirements.

Information resilience is stronger when it is layered across custody, protocols, access routes, and privacy rather than concentrated in one gatekeeper.

The report bodies include technical circumvention detail; public synthesis intentionally stays at the architecture/governance level.

The Russian model described by the source corpus combines legal, network, platform, surveillance, and punitive layers rather than one firewall.

Current details are time-sensitive and should be freshly verified before being treated as a present-tense implementation inventory.