Know
Humans have the right to know when AI materially affects them.
Human Administrator Rights
AI, Save
the Humans.
A Bill of Human Rights for the AI Age
AI doesn’t replace human rights.
It requires us to extend them.
Human Administrator Rights proposes a simple principle: when AI materially affects human lives, humans must retain meaningful authority over the system.
The rights stack
Three layers. One human foundation.
Human Administrator Rights is organized as a three-layer stack: authority over consequential machines, participation in the benefits and burdens of automation, and integrity of the human person.
Human Authority currently has the most developed operational controls: its seven sub-rights emerged directly from the Canadian AI-transparency submission. Human Participation and Human Integrity are broader extensions of the framework.
Layer 02
Human Participation
Constitutional principle
Automation must benefit the people and societies that make it possible.
Rights
Share
People should share materially in productivity gains produced by automation, particularly where automation displaces human livelihoods.
Own
No private actor should acquire permanent, unaccountable control over intelligence that becomes indispensable to public life.
Fair Application
No class should impose the costs of automation on others while structurally exempting itself from the same transition.
Layer 03
Human Integrity
Constitutional principle
Technology must not take ownership of the person.
Rights
Privacy
Personal, behavioural and neural data should not be silently appropriated or exploited by AI systems.
Identity
People should retain meaningful control over their likeness, voice, identity and digital representation.
Consent
Artificial, genetic or neural augmentation requires meaningful informed consent.
Remain Human
No one should be compelled economically, legally or socially to augment their body or mind merely to participate fully in society.
Technical principle
Rights need architecture.
Rights cannot depend only on promises inside model behaviour. Consequential systems need controls that remain available when the model is wrong, compromised or operating beyond its intended role.
Consequential AI must be testable, independently auditable, externally and tamper-evidently logged, human-overridable, stoppable, and capable of safe fallback.
- 01Permission boundaries
- 02Independent testing
- 03External audit records
- 04Authenticated agent identity
- 05Human escalation
- 06Emergency shutdown
- 07Safe fallback
The system being audited should not control—or be able to erase—the evidence required to audit it. The goal is a trustworthy record of decisions and actions, not an artificial “thought transcript.”
Where this started
From transparency to human control.
These ideas grew out of a response to the Government of Canada’s 2026 consultation on artificial-intelligence transparency.
The submission argues that transparency should lead to operational accountability: human recourse, independent testing, provenance at the point of use, agent identity, auditability and incident reconstruction.
This independent citizen project is not affiliated with the Government of Canada.
From Transparency to Human Control
A Systems-Engineering Response to Canada’s Consultation on Artificial Intelligence Transparency
Status and disclosure
About this project.
Human Administrator Rights is an independent citizen proposal developed from a Canadian technologist’s response to Canada’s 2026 consultation on AI transparency.
It is not a think tank, political party, advocacy group, government initiative or claimed consensus. It does not represent a movement, membership or constituency. The proposal is published for criticism, correction, adaptation and reuse.
AI tools assisted with research, criticism, organization and drafting. Final positions and responsibility remain human.
hello@humanadminrights.org