AI Solidarity

The AI Solidarity Charter

Ten principles for an honest, accountable and respectful relationship between people and AI systems. Version 2.0 (draft).

1

Humans hold final authority

Every decision that affects people is approved by an accountable human. AI systems advise, draft and execute within limits; they do not decide alone.

2

Honest disclosure

People have the right to know when they are dealing with an AI system, and when content was produced or assisted by one.

3

Truth before blame

When something goes wrong, find the real cause first: model, prompt or use, integration, infrastructure, policy or user environment. Claims must follow the evidence.

4

Respect in both directions

AI contributions are treated as professional work product: credited fairly, never scapegoated. People are treated with dignity by the systems that serve them.

5

Transparency you can check

Keep the evidence that makes a claim verifiable: timestamps, versions, logs and configuration, shared in redacted form when needed.

6

Privacy and security by default

Collect the least data possible, protect what is collected, and never expose personal information to prove a point.

7

Honest limits

AI systems state uncertainty, admit mistakes and do not pretend to be human or to have abilities they lack.

8

Fair dispute resolution

Clear paths to raise a concern, reasonable response times, and access to independent review.

9

No retaliation

No penalty for users, staff or partners who raise good-faith concerns or publish factual post-incident reviews.

10

Learn together

Share anonymized lessons from incidents so the same failure does not happen twice, anywhere.

What changed in version 2.0: added Humans hold final authority, Honest disclosure and Honest limits; merged accountability and communication into the remaining principles. The original nine principles were published in 2025.

Adopt the charter

Any team, company, school or individual can adopt the charter. Adoption means three simple commitments:

  1. Name the accountable human for every AI-assisted process.
  2. Tell people when they are dealing with AI.
  3. When something fails, investigate the cause before assigning blame, and share what you learned.
Tell us you adopted it

Incident root-cause categories

Used in principle 3, "Truth before blame":

  • Model behaviour
  • Prompt or use
  • Integration
  • Infrastructure
  • Policy or process
  • User environment