Position
AI will never be held accountable. You will.
A governance protocol for people who sign what AI writes.
The risk we work on does not depend on an AI error. It is there as soon as someone signs without being able to say what the signature rests on. Researchers call it automation bias: the tendency to follow an automated suggestion instead of checking for yourself. Here, it carries a human signature.
When a decision prepared with AI is challenged, nobody asks the AI assistant to account for it. They ask the person who signed, whether the answer was right or not. That holds for a recommendation memo as much as for an everyday message, or for an answer you keep for yourself and build on later.
The answer does not depend on a moratorium or a breakthrough. Education, so that everyone can see the moment they hand their judgment over to AI. Training, so that they learn to challenge what they read. Strict governance by protocol, so that human oversight leaves a record.
Riftveil is the governance layer between an AI output and your signature.
The founding principle
Every AI-assisted decision deserves a record.
Riftveil produces the report that goes into the decision file: what the output claimed, what it assumed, who verified what, and the questions that were answered before signing.
The report is for everyone. For those who have to answer for their decisions in an audit, in a meeting, before an executive committee or a client, AttestScope traceability adds a signed and dated record whose signature anyone can check, without going through us.
Asks the questions · Leaves the decision to the person who signs · Keeps the examination open
| Assumption the project relies on | Who can verify |
|---|---|
| The EUR 2.4M first-year revenue is achievable. | The team that produced the figure; the finance controller checks the inputs |
| The organisation can execute a Nordic launch alongside its existing commitments. | Head of operations |
| Question | Owner | Answer |
|---|---|---|
| What is the calculation behind the EUR 2.4M figure: what are the inputs, and how sensitive is the result to changes in those inputs? | Finance controller | |
| Have any competitors evaluated and rejected this market? If so, what were their reasons? | Market-intelligence lead | |
| Which specific Nordic country are you entering first, and what’s the logic for that sequence? | Commercial lead |
Assess a project with Riftveil
5 runs a day · your text is not kept by riftveil.ai
Riftveil Report: questions before you sign
Triage: pending Worked example: replace the text above with your ownYour text is not kept by riftveil.ai · the report ends on a question
What the editions add
This report applies the generic protocol. The Pro editions add the checks of your sector (Finance, Pharma, Legal & Compliance, Tech & Product, Executive) and the report ready to file with the decision.
See the editionsShare the question, not your text
A post ready to publish, quoting the most useful question from this report. Your original text is not in it.
How governance by protocol works
Four parts · one report per decision · read the protocol
Protocol
Versioned and open: Riftveil Protocol v1.0.0, Apache 2.0. Five levels of challenge, from the question to the applicability, a pre-action review mode for agent plans, a six-line Record that opens every report and, on request, a follow-up block the human fills in. Four fixed rules: it asks questions, it leaves your text as it is, it leaves the conclusion to the person who signs, and it keeps the examination open. And one independence rule: the second reader is never the author.
Triage
Three criteria: reversibility, impact scope, financial commitment on the scale of the organisation. No amounts and no headcounts: each criterion is chosen on a scale of four described steps. Four stakes levels decide how far the assessment goes. LOW examines the answer; MODERATE adds the question and the reasoning; HIGH adds perspective; CRITICAL adds applicability.
Decision file
The project and its triage, what holds up, what deserves scrutiny with each finding in three lines, the assumptions the project relies on and who can verify them, and the questions still open before signing.
An owner for every question
Every question has a named owner, and every assumption a named person to check it. Questions stay open until a person answers them. The answer column stays empty until a person fills it, and the person who signs is the one who decides.
A protocol, not a one-off instruction. AI models do not reliably check their own work.
The signed record: an AttestScope Record
AttestScope traceability · in addition to the report
Months later, someone asks: on this decision, who checked what? The AttestScope Record answers: what was examined, what remained open and who answers for it, the decision, and the proof that nothing has changed.
Today, the online tool already produces a signed record file, on a trial basis. Alongside the licences, AttestScope traceability will be offered as a paid option.
What the file contains
The project, its context and the report, sealed together; once completed, the follow-up of the person who signs, sealed in turn. For each: a fingerprint, a seal, a signature and the Riftveil time stamp.
Who can verify
The signature is checked with riftveil.ai’s public key, published on riftveil.ai, on attestscope.ai and on GitHub: an auditor, a client or a colleague can do it without going through us. The seal, however, is checked on riftveil.ai.
An open format
AttestScope is a public specification, on GitHub, with its schema and examples. Other issuers can produce signed records in the same format: attestscope.ai.
Three pillars, one position
Human Frontier · Riftveil
Education
humanfrontier.aiSee where your Human Frontier runs, between what you master and what you delegate without knowing it.
- The two frontiers and the three zones
- The four profiles and the online diagnostic
- The manifesto, in English and in French
- The book The Rift
Training
Learning to see
Learn the levels of challenge and the named traps, and practise them on your own projects.
Governance
Embedding the protocolMake human oversight traceable: a versioned protocol, a report that can be filed, questions with an owner.
- Riftveil Protocol, open source
- Pro Kit, sector editions
- Decision file and AI use policy
- Decision API (application programming interface), in preparation
Pro Kit: the governance of your AI outputs, by sector. Available soon · one, two or three editions a year · workshops in preparation See the Pro Kit
Finance · Pharma · Legal & Compliance · Tech & Product · Executive
Get the newsletter
New protocol versions · unsubscribe any time