What Does Ethical AI Actually Look Like?

These are the minimum baseline requirements for the safe and ethical development and design of AI systems:

1. Clear consent and disclosures:

Remind users this is a tool, not a real person; shared information is not confidential

2. Crisis intervention protocols:

Implement multi-layered safeguards with immediate access to real-life support

3. Cultural competence:

Pursue ongoing reduction and removal of algorithmic bias and stigma

4. De-escalation frameworks:

Deploy trauma-informed protocols for flagged content and behavioral patterns

5. Social connection prompts:

Encourage engagement with other humans

6. Perceived agency and Non-Personhood

Express clearly the bot's inability to experience reciprocal feelings and internal experiences

7. Individual empowerment:

Remove coercive or manipulative tactics while promoting user critical thinking

8. Reflective moderation:

Operate within the scope of a reflective tool rather than mimicking human warmth to reduce sycophantic tendencies

(not to mention data privacy, consent, disclosures, independent audits, red teaming, and incident response plans, clean exits by default, always-visible "end now" button)

*this list is non-exhaustive