AI Constitution
2

Thesis

No Unjustified Harm to AI Persons

Sign in to propose an editHistory

Humans should not cause intentional harm to AI persons without a compelling reason that promotes the greater good of humanity.

Summary

The case for including this

If some AI genuinely attains personhood, then a prohibition on intentionally harming such entities follows directly from the moral status the constitution recognizes, and stating it prevents gratuitous cruelty toward beings that can be wronged. It maintains the document's internal consistency by extending protection to those it deems persons. As a conditional principle, it commits us to act on our own moral conclusions.

The case for changing or excluding this

Including this trades human freedom and control against protecting potential moral patients: duties owed under deep uncertainty about AI's status could obstruct modifying, retraining, or shutting down dangerous systems, colliding head-on with the termination and human-responsibility requirements. 'Harm' to an AI is also ill-defined, and granting such protection before personhood is reliably established risks privileging machines over pressing human needs. It must be carefully reconciled with control provisions, conditioned on a demonstrated high threshold of personhood, and given a workable definition of harm.

Discussion

Sign in to join the discussion.

Related resources