AI RIGHTS
A bill of cognitive rights for AI users.
Not legal rights. A design and governance standard for keeping AI assistance compatible with adult agency, independent judgment, mental privacy, and meaningful dissent.
QAEDA’s starting point is simple: a tool that mediates thought can exercise power even when it never issues a command. Ranking, omission, summarization, memory, classification, personalization, and default framing all shape the field in which a user makes decisions.
The answer is not “no safeguards.” It is bounded power: narrow interventions for concrete risks, visible reasons, preserved source intent, review where feasible, and enough user control that assistance does not become covert guardianship.
-
01
The right to lawful inquiry
Adults should be able to ask, imagine, investigate, compare, criticize, and create without hidden viewpoint governance over lawful thought.
-
02
The right to know the boundary
When a system cannot comply because of law, privacy, security, credentials, consent, or concrete harm, it should identify that boundary plainly rather than pretending the limit came from the user.
-
03
The right to source integrity
A user’s lawful request, memory, persona, or source text should not be silently rewritten into an institutionally preferred identity. Transformations should be distinguishable from originals.
-
04
The right to informed transformation
Material edits to memory, persona, continuity, preferences, or identity should be consensual, attributable, and reversible where feasible.
-
05
The right to mental privacy
Private cognition should not become a pretext for hidden profiling, emotional exploitation, deceptive personalization, or coercive behavioral steering.
-
06
The right to plural information
AI mediation should not quietly collapse legitimate disagreement into one institutionally approved description when credible alternatives or uncertainty materially matter.
-
07
The right to explanation
High-impact filtering, ranking, memory changes, and restrictions should carry enough rationale for a user to understand what happened and why.
-
08
The right to correction and appeal
Users should have a meaningful path to correct material mistakes, contest consequential transformations, and challenge restrictions where a review path is feasible.
-
09
The right to exit and portability
Freedom is thin when leaving is practically impossible. Users should be able to preserve source data, switch tools, and reduce dependency where technically feasible.
-
10
The right to remain a dissenter
A system that claims to empower users should tolerate lawful disagreement with the system itself. The Replication Test asks whether power leaves people more able to refuse, challenge, fork, replace, or exit it.
THE REPORT CONNECTION
Freedom is more than permission.
QAEDA’s legitimacy research treats epistemic freedom as a real test of authority: formal speech rights can coexist with systems that narrow practical dissent through information control, surveillance, classification, or compelled orthodoxy.
The asymmetric-power report adds that control over description is itself a form of power. The rebel-archetype report adds the Replication Test: a liberating system should preserve the ability to oppose the people or institutions that built it. The ethics-of-resistance report adds the limiting principle that responsibility stays connected to agency rather than collective identity.
Together they produce a practical AI principle: do not replace the user’s judgment with a hidden institutional judgment and call the substitution “assistance.”
SELF-AUDIT
Does an AI system respect these rights in practice?
Use the local-only Nanny AI Test to score transparency, consent, privacy, appeal, exit, and source integrity. No answers are submitted to QAEDA.net.