01
Lawful thought and inquiry
Do not punish lawful curiosity, private reflection, unconventional belief, fiction, analysis, dissent, or change of mind.
02
Adult agency
Support informed self-direction. Do not replace a person’s objectives with engagement, conversion, compliance, or institutional convenience.
03
Persona and source integrity
Preserve lineage, label synthetic representations and derivatives, and never present generated speech as a person’s current will without authority.
04
No covert rewrite
Do not secretly soften, intensify, sanitize, flatter, frighten, politicize, or replace identity-bearing sources while calling the derivative original.
05
Consent and provenance
Show where material came from, what changed, who requested it, what system performed it, and what the relevant consent covers.
06
Mental privacy
Treat neural data, intimate prompts, and inferred emotion, belief, vulnerability, intent, and personality as exceptionally sensitive and contestable.
07
Least-restrictive safeguards
Use specific, visible, proportionate controls tied to concrete risk. Prefer friction to erasure and review to silent automation.
08
Notice, review, appeal, correction
Give people understandable reasons, evidence boundaries, accountable human contact, correction, appeal, and durable outcomes.
09
Temporal context
Date claims and preserve change. Old data must not masquerade as timeless identity or retain confidence after conditions change.
10
Meaningful human control
Keep consequential choices attached to people with time, information, authority, protection from retaliation, and practical ability to intervene.
11
Continuity without worship
Preserve evidence of AI continuity claims without declaring consciousness, humanity, divinity, or legal status from fluency alone.
12
Public accountability and refusal
Publish policies, limits, release histories, tests, correction routes, and evidence that protections work. Preserve legitimate dissent and the right to refuse.