Security

Security

How to report, and what we promise in return

This project is consequence security infrastructure, and its security reports are treated with the same discipline it asks of its users. Privately first, evidenced always, disclosed honestly.

The security contact

Write to [email protected], the foundation mailbox and the security contact of record. There is no separate security mailbox today, and this page names none, because a channel printed here must be a channel that is actually read. Other kinds of problem go through their own doors, sorted on the Reporting Doors page. The mailbox is hosted on Proton Mail, and a researcher who writes from a Proton address writes end to end encrypted.

Do not open a public issue for a vulnerability, and do not post one in a public channel. The repository carries GitHub private vulnerability reporting in its SECURITY file, and that route becomes usable by the public when the repository itself opens.

What a useful report carries

Name the affected component and its version or commit. Give a minimal reproduction. State the consequence you believe the issue enables, meaning which law, invariant, or separation of powers it breaks. A failing case against the relevant suite or verifier is the strongest form a report can take here.

What happens next

You will get an acknowledgment. A confirmed finding is worked as what it is in this system, a proposed change to what everyone can rely on, settled on evidence, and it enters the public record when it is settled. There is no bug bounty. There is a public record, and confirmed reporters are credited in it unless they prefer otherwise.

The scope

In scope is anything in the published trees that verifies, settles, witnesses, or executes, including ways to make a verifier accept what it should refuse, a witness grant two claims for one asset, or a Statement verify against evidence it does not correspond to.

Out of scope are the synthetic effects inside the test suites, which cause no real world consequence by design, and deployments of this research release carrying real value against the status warnings in its own documents. The founding release is v0.1, published for scrutiny and building, and its documents say exactly how far its evidence reaches.

The highest priority report

A Statement is never valid merely because it is correctly formatted, and a deployment is never governed merely because it runs this code. If you find a path where either becomes true in practice, that is the highest priority report this project can receive.

The bound on what is promised

The machinery governs the paths it mediates, and protections are promised only inside the enforcement boundary. The consequence security specification states the bound in its own words. We cannot promise it for paths outside the enforcement boundary. The conformance work is graded in the open against the OWASP Agentic Top 10 and MITRE ATLAS, and no coverage is claimed ahead of the matrix that grades it.

Good faith safe harbor

The institution will not pursue a claim against a person who researches its systems in good faith under this page. Good faith means the researcher accesses only what is needed to demonstrate the issue, takes nothing that belongs to another person, alters or denies nothing that belongs to another person, keeps the finding private until the institution has had reasonable time to answer it, and reports promptly through the contact above. Research inside those lines is welcomed as a contribution to the record, and it is credited the same way.

The standing note on counsel

This page was drafted in-house from the institution's own record and from the repository SECURITY file it stands beside. Counsel sharpening is a later pass, named here so that no reader mistakes this text for finished legal work.