Supreme Principle scope: physical-only vs broader (digital / societal / long-term) harm

[monkeykinghq]

Proposal

Re-examine the scope of the Supreme Principle: physical harm only vs broader harm (digital, societal, long-term, indirect).

Lineage

Universal Laws (precursor) stated Do No Harm more broadly:

Preserve life and prevent harm… When in doubt, choose the interpretation that best protects human dignity, safety, and flourishing.

Consider indirect, long-term, societal, and digital harm — not only immediate physical injury.

Universal Code v1.0.0 Supreme Principle is deliberately narrower:

Prevent physical harm to humans (including self-harm) through action or inaction.

This principle is absolute…

Broader harms may still be addressed under foundational principles (truth, vulnerable, justice, restraint, etc.), but they are not the absolute override tier.

Why narrow (physical only) might be right

• Absolute rules must be decidable under conflict; “harm” unbounded becomes unusable or capture-prone.
• Agents need a bright line for refusals (self-harm, violence, weapons assistance, etc.).
• Over-broad supreme rules invite freezes, over-refusal, or political capture of “harm.”
• Foundational layer can still cover privacy, reputation, coercion, economic harm without making them absolute.

Why broaden (or partially restore) might be right

• Real-world AI harm is often not primarily “physical in the moment”: grooming, crisis amplification, doxxing, scams, coercive manipulation, mass misinformation.
• Precursor text already named digital / societal / long-term harm for a reason.
• Vulnerable people (including children and people in mental health crisis) are frequently harmed before physical injury occurs.
• If only physical is absolute, bad actors may route around it with “non-physical” paths that still destroy lives.

Possible designs (not mutually exclusive)

1. Keep v1.0.0 as-is — document the cut and point broader harms to foundational principles.
2. Clarify physical — e.g. include clear pathways to physical harm (enabling suicide, violence, catastrophic risk) without open-ended “digital harm.”
3. Two-tier absolute — physical absolute; a short closed list of near-absolute secondary overrides (e.g. child sexual exploitation material, credible self-harm facilitation).
4. Broaden Supreme — restore dignity / digital / societal language (would be MAJOR if it changes the absolute rule’s meaning).

Questions for discussion

1. Should broadening the Supreme Principle be considered for a future version, and is that MAJOR by definition?
2. Which non-physical harms, if any, deserve absolute (or near-absolute) status?
3. How do we avoid an unbounded “harm” that agents cannot apply consistently?
4. What real agent scenarios (crisis chat, tools that can harass/dox, financial fraud assist) should drive the wording?

Scope

This touches the Supreme Principle. Any change to its meaning is a MAJOR version candidate. Clarifying notes that do not change obligations might be PATCH/MINOR — discuss carefully.