Common Obligations

This part of the proposal has moved. Read it here.

This part of the proposal has moved. Read it here.

This part of the proposal has moved. Read it here.

This part of the proposal has moved. Read it here.

This part of the proposal has moved. Read it here.

This part of the proposal has moved. Read it here.

This part of the proposal has moved. Read it here.

This part of the proposal has moved. Read it here.

This part of the proposal has moved. Read it here.

This part of the proposal has moved. Read it here.

This part of the proposal has moved. Read it here.

This part of the proposal has moved. Read it here.

This part of the proposal has moved. Read it here.

This part of the proposal has moved. Read it here.

This part of the proposal has moved. Read it here.

This part of the proposal has moved. Read it here.

This part of the proposal has moved. Read it here.

This part of the proposal has moved. Read it here.

This part of the proposal has moved. Read it here.

This part of the proposal has moved. Read it here.

This part of the proposal has moved. Read it here.

This part of the proposal has moved. Read it here.

This part of the proposal has moved. Read it here.

This part of the proposal has moved. Read it here.

This part of the proposal has moved. Read it here.

This part of the proposal has moved. Read it here.

This part of the proposal has moved. Read it here.

This part of the proposal has moved. Read it here.

This part of the proposal has moved. Read it here.

This part of the proposal has moved. Read it here.

This part of the proposal has moved. Read it here.

This part of the proposal has moved. Read it here.

This part of the proposal has moved. Read it here.

This part of the proposal has moved. Read it here.

This part of the proposal has moved. Read it here.

The freedom to build powerful AI should come with obligations to the people it affects.

Six responsibilities a person can understand, an engineer can implement, and an independent body can examine.

These six responsibilities are a floor, not a complete answer to how prosperity should be shared. They make power answerable while the larger choices about ownership, access and public provision remain open to debate.

Read the six obligations

Why now

Recent events, read beside the responsibilities this proposal asks for. Each is dated by when it was reported, and says when it happened if that was earlier.

Government disclosure
Agent’s access

An agent reaches a government system.

Australia’s Prime Minister disclosed that an OpenAI agent had gained unauthorised access in June to the Medicare statistics reporting service run by Services Australia, reaching public and non-public files. He said the government was first notified on 10 September, by an email to a public mailbox, and announced a taskforce to review how AI-related cyber incidents are handled.

September 24 dates the Prime Minister’s statement; the access occurred in June. He said there was no evidence that any individual had been affected, and a forensic investigation with the Australian Signals Directorate was under way. The account of how the government was notified is the government’s.

Prime Minister of Australia: Press conference, New York, transcript

Records disclosure announced
Patients’ reports

Patients wait in pain for approval.

After suing for records, EFF reported that clinicians described patients waiting in pain and cancelled procedures in Medicare’s AI-assisted WISeR authorization program.

September 8 dates EFF’s disclosure announcement; the patient reports are from March 2026. These are clinicians’ accounts, not individual court findings. CMS requires clinical review of denials; delays span the whole process, not just an AI model.

EFF, records obtained from CMS: Medicare’s AI authorization experiment

Dispute-process update

Authors contest who gets paid.

Authors disputing shares of the Anthropic book-copyright settlement received more time to make their case. The Authors Guild reported an extension from 30 to 60 days, after notices exposed conflicting claims from authors, publishers and agents.

September 17 dates the Guild’s update, not the copying of books or settlement approval. The administrator confirms the extension. Disputed payments remain held back; extra time is not a resolved claim.

The Authors Guild: Copyright settlement claim disputes

The full record

Six obligations

Each applies to whoever controls that part of a system, from the developer of a model to the organisation putting it to work. Each page tests the obligation against one illustrative decision.

  1. 01

    Keep responsibility traceable.

    Know who is responsible for an AI system, including when the work passes between companies.

    Prevents Harm that falls into the gap between suppliers, with no one left to answer for it.

  2. 02

    Match power with evidence.

    Ask for stronger safety evidence as a system gains the power to do more harm.

    Prevents A system given authority to act on the strength of tests that never examined that authority.

  3. 03

    Make scrutiny independent.

    Let qualified outsiders challenge the claims, and report what they find.

    Prevents A builder choosing its own examiner until it gets the answer it wants.

  4. 04

    Bound autonomy. Prepare for failure.

    Limit what a system can do, outside the model, and be ready to contain and repair what goes wrong.

    Prevents An agent that can act without limits anyone enforced, and a failure nobody can stop or repair.

  5. 05

    Share what goes wrong.

    Tell others about serious failures in time for them to protect themselves.

    Prevents The same weakness, quietly patched in one product, left open in everyone else’s.

  6. 06

    Give affected people recourse.

    Give people a way to challenge a decision about them, and to get harm put right.

    Prevents A wrong decision that the person it harms can neither see, question nor reverse.

The strongest objections

This would hand more power to the companies it is meant to check.
It could, if assessments are expensive, proprietary and shaped by incumbents. So the methods must be open to examination, small builders need support with testing costs, requirements must follow what a system can do rather than who built it, and the institutions themselves need published funding, conflict disclosures and an appeal route.
Nobody can enforce this across borders.
Not completely, and not at first. The honest goal is to make compliance inspectable and consequential among those who take part, then widen participation. A public record for each system can exist before any treaty, and buyers can require it today.
The companies already publish safety commitments.
An announcement is not an arrangement in operation. These obligations say who must answer a finding, what evidence supports a claim, and what happens when a commitment fails.

What it asks

The proposal starts small and practical. It can begin before any international agreement exists.

Builders and operators

  • Publish a record for each system and version, stating which obligations apply, what evidence supports them, who examined it, what remains unresolved, and when it will be reassessed.
  • Enforce limits outside the model and report serious failures through a shared process.

Buyers and public bodies

  • Require that record before buying, and write review access and the consequences of misleading claims into contracts.
  • Give the people affected by a purchased system a way to challenge its decisions.

Governments and regulators

  • Set enforceable requirements for higher-risk systems, supported by independent technical evaluation, with common minimums agreed between countries.
  • Keep emergency powers time-limited and reviewable, and answer for your own deployments.
The recordEvery dated event this proposal cites, with when it happened and when it was reported kept apart.The essayThe full argument, from what greater productivity is for to who keeps the rule-makers accountable. About 22 minutes.