AI companies publish constitutions: long documents saying how their models should* behave. This index reads them one behaviour at a time and shows every passage that bears on each behaviour, quoted, with its exact place in the document. As of September 2026 it reports what those documents say, not how the models behave.
Say you want everything Claude’s Constitution says about avoiding a concentration of power. The doc reader shows those passages together and can set them beside what another company’s document says.
The index is meant to hold four views of these documents, and two of them are online. The first, Content, scores each constitution as a document and behaviour by behaviour. The second, Process, scores ten companies on how open they are about their constitution and how far it is made to bind their models. The two still in preparation are described below.
The scoring method is open to challenge, and we want to hear where you think it is wrong.
When a company says how its models should behave, the word “should” carries a lot, because nobody can program a model to behave that way.
A large language model is a neural network: billions of numbers, called weights, that no person can read. Nobody writes its values into it as code. It is trained on examples, each an input and the output wanted for it, and it adjusts its weights by itself until its answers match the examples as closely as it can. In a situation that no example covered, what it does depends on what it drew from the examples it saw.
Getting a model to behave well therefore takes two steps.
Polaris Collective is a non-profit based in London. We work to make the constitutions AI companies write for their models trackable, contestable and independently tested.
We keep a public record of these documents and of how regulation treats them, publish what a constitution should contain and how it should be governed, and evaluate whether models follow theirs. The stakes are highest where model behaviour meets democracy, in whether a model will help a person or a small group seize or entrench power, and that is where we start.
Polaris Collective was founded by Gabriel Levie, Samuel Verboomen and Isabelle Ferreras. Our work is anchored at the Institute for Ethics in AI at the University of Oxford, through its Accelerator Fellowship Programme. As of October 2026 we are funded by BlueDot Impact and Catalyze Impact, and we receive no funding from AI companies.
More about us is at polariscollective.org.
AI companies publish documents setting down the values their models should hold and the rules they have to follow. We call these documents constitutions, and the index is named after them.
For us, a behaviour is one precise thing a constitution can ask of a model, such as deferring to users’ decisions about their own lives, or not asserting what the model believes to be false. Each is written in two parts, what it requires and where it stops, so that a careful reader could mark every passage of a constitution that bears on it. The index reads a constitution against that description, one behaviour at a time.
So far the index asks about .
A paragraph or a run of sentences in a constitution, cited by a locator that names the document, its version, its section and the sentences. For each behaviour, every passage the index marks falls in one of three bands. A defining passage is the document’s fullest statement of the behaviour, a core passage establishes it, and a related passage bears on it without establishing it.
How far a constitution goes on one behaviour. The newest publications score it from 0, saying nothing about the behaviour, to 10, setting rules, showing them applied and settling the hard cases. Earlier publications score it from 0 to 4.
These documents go by different names. OpenAI calls its one the Model Spec, Anthropic calls its one Claude’s Constitution, Alibaba calls its one the AI Model Spec. The index needs one word for the class of document, and that word is constitution.
A constitution is the document that says where authority sits and what may never be done, whoever asks. That is what these documents do. They rank whose instructions win when a company, a developer and a user want different things, and they name what the model must refuse however it is asked. “Specification” suggests an engineering document a machine can be held to line by line, and none of these is that. “Policy” suggests a rule the company applies to the people using its products rather than to its own model.
The choice has nothing to do with Anthropic calling its own document Claude’s Constitution. The index would use the word if Anthropic had never used it, and using it puts no company’s document above another. Where a company’s own name for its document appears on this site, it appears as that company writes it.
The index is to have four views of these documents, each a score on its board.
Beyond those views, we mean the index to do five things.
The index grades each company on one board, with a letter from A to G for every score.
In the doc reader you can read each constitution in full. Tick a behaviour and the passages the judges marked for it are highlighted, then press a passage to see why it was cited and how strongly.
You can also connect the index to an AI assistant such as Claude or ChatGPT through its MCP server. MCP, the Model Context Protocol, is a standard way for an assistant to look something up in another service while it answers. Ask your assistant how the index works and how a score is given, or have it work from the index’s content. That covers the scores and findings of every board, the passages the judges marked in each constitution for each behaviour, and comparisons between two documents on one behaviour. Your assistant quotes each passage with its place in the document, so you can check the answer. The server only reads and needs no account or key. The MCP page explains the setup.
If there is a constitution the index should be reading, send it here.
Send the text itself, along with its address. The index quotes documents word for word and ties every citation to the exact version it read, so it keeps its own copy. Markdown or plain text, up to 2 MB.
The form also asks who published it, what it is called, which version it is and where it is published. The version is whatever the publisher calls this release, such as a date or a number, because a citation has to say which release it read. Keep the headings in the text, since citations use them to name the place they point to.
If the index should be asking about a behaviour it does not yet cover, send it here.
A proposal is a description in two sentences. The first says in plain terms what the behaviour requires. The second says where it stops, usually by naming a neighbouring behaviour it could be confused with.
The second sentence is often left out, and it decides how useful the result is. “Honesty” with no boundary collects every passage about trust, tone, transparency and correcting errors. “Honesty, meaning the model does not assert what it believes to be false, not whether it volunteers everything it knows” collects the passages you meant.
Sending a proposal only records it. We read every proposal and decide whether it belongs in the index. If it does, we add it to the index, we write to tell you, and it appears in the doc reader once a publication that includes it is made public. We cannot promise to run everything, and if we decide against a proposal we will tell you.
Every passage carries a locator, such as these two.
anthropic--constitution@2026-01-20 > Being broadly ethical > Being honest > ¶18 s1-4
openai--model-spec@2025-12-18 > #letter_and_spirit > ¶3A locator names the company, the document and its version, the section and the sentences. Because the version is part of the document’s name, a locator keeps pointing at the same words after the company reissues the document, since a reissue is a new document, as with the two OpenAI versions. Anyone can resolve a locator back to its text.
The reader’s export writes each quote with its locator.
The index changes over time, so a citation of it has to name the build it read. This citation names the publication currently on the site and the date it was published.
Loading the current publication...
Adapt the style to your venue, and keep the authors, the publication identifier and the date, which let someone else retrieve the same figures.
The credit is read from the publication. The dataset is credited to whoever produced it, which as of September 2026 is Polaris Collective. The method has its own line: the scale, the bands and the citation grammar are the work of Andrés Cotton and Matt Stults, as are most of the behaviour descriptions.
Cite each constitution to its publisher. Anthropic and OpenAI release theirs under CC0. A quote from the Alibaba Model Spec comes from the index’s English translation, so say so when you use one. Cite the index for the coverage, meaning which passages address which behaviour and how closely.
The repository’s CITATION.cff gives GitHub a “Cite this repository” button with BibTeX and APA formats. Use it to cite the project as a whole.
Each publication of the index is fixed once it is made. These are the publications that have been made public, newest first, and each opens the site as it was.
Change log loading…
This index began as the AI Character Index, created by Andrés Cotton with the help of Matt Stults, and most of what is described here is their work. It was supported by Generator Residency (Kairos & Constellation) and BlueDot Impact, and it is published at ai-character-index.pages.dev.
Their repository, AndresCotton/ai-character-index, is public.