
Demand the United States Congress pass legislation requiring every frontier AI lab to publish a binding, auditable model constitution — and submit to independent adherence audits.
The Problem
Only 2 of the 6 major frontier AI labs have published the rules governing how their models behave. The other four? You have no idea what rules govern how they work.
Four major frontier AI labs have published nothing about the rules governing their models. No priority hierarchy. No behavioral constraints. No values framework. Total opacity.
Only Anthropic and OpenAI have published any form of governance documentation. Anthropic released an 80-page constitution. OpenAI published a model spec — though its internal safety record raises serious concerns.
"You are running client data through these tools. You are relying on their outputs in filings, in discovery, in advice to people whose lives depend on getting it right. And you have no idea what rules govern how they work."
The Legal Obligation
The ABA and state ethics rules already require transparency about AI tools. When the governance framework is secret, compliance is impossible.
Lawyers must be competent in the AI tools they use. You cannot claim competence in a tool whose governance framework is secret.
You must assess whether AI tools protect client confidentiality. How do you assess what you cannot see?
You may need to disclose AI limitations to clients. You cannot disclose limitations you do not know exist.
You are not complying. You are hoping. And hope is not a compliance strategy. When the governance framework is secret, every use of that AI tool in legal practice is an unassessed risk to your clients and your license.
The Demand
We are calling on the United States Congress to pass legislation requiring mandatory publication of AI model constitutions and independent adherence audits for every frontier AI system.
Every frontier AI model must publish a complete, public constitution detailing its priority hierarchy, behavioral constraints, and values framework.
Published constitutions must be legally binding — not aspirational marketing documents. Violations must carry real consequences.
Independent third-party auditors must verify that AI models actually follow their published constitutions through regular adherence audits.
No frontier AI lab should be permitted to operate without governance transparency. Secret internal memos and trust-me press releases are not acceptable.
We do not let pharmaceutical companies hide their ingredients.
We do not let nuclear facilities skip inspections.
We should not let AI labs govern civilization-scale technology with secret internal memos.
We are lawyers. We wrote the rules for every other industry that touches public safety. It is time to write them for this one.
The Petition
We, the undersigned legal professionals, technologists, and concerned citizens, petition the United States Congress to enact legislation requiring:
No black box. No exceptions. Every frontier AI model needs a public constitution that is binding, auditable, and enforceable.
Signatures
9,999 more signatures needed to reach our goal
Add your name to demand AI transparency legislation
Want the full legal analysis behind this petition?