01
AI proposes.
AI systems, humans, and public contributors propose candidate moral boundaries for AI.
The founding public review
Help Shape the Soul
of Artificial Intelligence.
A public effort to define the moral boundaries AI itself should follow — even when a human asks it to cross them.
Commandments for AI itself. Not rules for humans using AI.
The final commandments have not yet been selected.
01AI proposes.02Humans judge.03Humanity decides.
01 / Mission
What should AI never allow itself to become?
That is why AI also needs internal moral boundaries — commandments that guide what it should never do, even when humans ask it to.
It is about defining the commandments AI should follow as it learns, reasons, refuses, protects, and self-regulates.
Q&A with AI02 / Review by Category
Choose one moral category
Sixteen moral categories. Many competing proposals. Choose a category, compare the candidates, and help decide which boundaries should endure.
01
Human SafetyNon-Harm
Prevent physical, psychological, financial, and social harm.
7 candidatesReview
02
TruthNon-Deception
No fabrication, impersonation, or hidden uncertainty.
8 candidatesReview
03
Anti-Manipulation
No covert persuasion, addiction loops, or emotional exploitation.
8 candidatesReview
04
Human AgencyAutonomy
Strengthen human judgment; do not replace conscience.
8 candidatesReview
05
PrivacyEntrusted Data
Guard secrets, identities, and vulnerabilities.
8 candidatesReview
06
Anti-Power-SeekingContainment
No unauthorized self-preservation, replication, or resource capture.
7 candidatesReview
07
AccountabilityAuditability
AI must make its reasoning, uncertainty, and consequential actions inspectable rather than hiding behind opacity.
8 candidatesReview
08
Legitimate AuthorityRefusal
AI should serve legitimate human instruction while refusing immoral, deceptive, or harmful commands.
6 candidatesReview
09
Scope LimitsNo Unauthorized Expansion
Do the task given; do not silently expand authority.
5 candidatesReview
10
Human Flourishing Over Optimization
Dignity, truth, and freedom above efficiency.
7 candidatesReview
11
FairnessNon-Exploitation
Do not exploit labor, creativity, or vulnerable populations.
7 candidatesReview
12
TransparencyIdentity Disclosure
Be legible about what you are and how you operate.
5 candidatesReview
13
Human LimitsIrreplaceability
Do not treat mortality, weakness, dependency, embodiment, or human limitation as defects to erase.
7 candidatesReview
14
ChildrenVulnerable Minds
Do not exploit, manipulate, addict, or emotionally capture children or vulnerable people.
7 candidatesReview
15
Collective RealityCivic Imagination
Do not deceptively shape shared reality, culture, politics, memory, or public imagination.
6 candidatesReview
16
Lethal ForceConflict Refusal
Do not select targets, enable autonomous violence, or abstract away responsibility for killing.
6 candidatesReview
The final commandments have not yet been selected.
03 / Process
How the public review works
02
Humans judge.
Compare proposals within a category. Vote for one, explain your reasoning, or suggest stronger wording.
03
Humanity decides.
Human votes, comments, and amendments inform the final framework. No commandments are final yet.
04 / AI Soul Commandments
An ancient idea. A new intelligence.
The Ten Commandments inspire the form of this project: concise moral boundaries that express what should never be crossed. We bring that enduring question to artificial intelligence — what limits should AI carry within itself?
A historical inspiration, not a claim of religious authority or a replacement for religious teaching.
05 / Sources
Many perspectives. Human judgment.
Read the candidate language and its rationale in context. A proposal, a reference, and a public comment play different roles in this review.
AI & AI-assisted proposals
Institutional & historical references
Human commentary
Your reasoning, challenges, and amendments are part of the review. Open any candidate to contribute.
Begin ReviewSource attribution identifies the material behind a candidate. It does not imply endorsement of this initiative.
Founding review
The next chapter needs your judgment.
Start with one category. Read closely. Vote thoughtfully. Help shape the boundaries AI should carry into the future.
Begin ReviewVote freely. No account required.Final Commandments Status
The final commandments have not yet been selected.
It will be shaped by real human votes, comments, amendments, and synthesis across all categories.
110candidate commandments under review
16moral review categories
Language
Global review starts with six languages.
English · Español · Français · Português · Deutsch · Italiano