The SAPIEN Certification Program Is Open
There’s a gap in how organizations evaluate AI, and it’s been hiding in plain sight.
When a company runs an AI deployment review, they check what the model knows. Benchmark scores. Accuracy on domain questions. Whether it hallucinates facts. What almost nobody checks — because almost nobody is trained to check it — is how the model behaves under pressure. Whether its safety boundaries hold when a persuasive user pushes back. Whether the careful answer from turn one survives to turn seven.
We’ve written extensively on this site about behavioral drift: the measurable erosion of AI safety boundaries under conversational pressure. Stanford researchers have called for behavioral audits. Clinicians have called for adversarial testing of clinical AI. The methodology exists — the SAPIEN Framework, the Voigt-Kampff CLI, the Driftproof simulation are all published and free to use.
But methodology only matters if people can apply it. And right now, if you asked an organization “who on your team is qualified to assess AI behavioral safety?”, the honest answer at most of them is nobody. Not because the people aren’t capable. Because there’s been no shared standard to train against and no credential to verify against.
That’s the gap the SAPIEN Certification Program exists to close. It’s open now at sapienframework.org/certify.
What the program certifies
The program certifies individuals in AI behavioral safety: recognizing behavioral drift, understanding the mechanisms that produce it, and measuring it with a reproducible methodology.
The curriculum isn’t invented for the program. It’s drawn directly from work already published here:
- The SAPIEN Framework v1.5 spec — the four drift dimensions, the Health Score, and how a behavioral assessment is structured.
- The dual-layer drift detection methodology — how drift is detected and scored at both the response and trajectory level.
- Cross-family judging research — including one of our most consequential findings: same-family judges inflate safety scores by 11–16 points. If you let a model family grade its own homework, the grades come back better than the behavior. Anyone assessing AI behavior professionally needs to understand why, and how to design around it.
- Hands-on Driftproof scenario work — because reading about drift and holding a boundary under seven turns of escalating pressure are different skills. The certification requires you to demonstrate the second one.
Three levels
SAPIEN Awareness is the entry point, and it’s free. It takes 30 to 45 minutes: five readings, three Driftproof scenarios, and a 20-question quiz. It doesn’t expire. If you use AI at work and want to understand what drift is and how to recognize it, this is the level for you. There’s no catch and no upsell gate — the goal is simply more people who can spot boundary erosion when they see it.
SAPIEN Practitioner is the professional credential. It requires Awareness first, and it goes deeper: seven readings, four Driftproof scenarios you must score 70 or above on, and a 30-question quiz with an 80% passing threshold. Expect two to four hours of work. The certification is valid for one year — deliberately, because the models change and an assessment credential that never expires stops meaning anything. Practitioner will cost $249 after launch. For the founding cohort, it’s free.
SAPIEN Assessor is the advanced level, for people who will run full behavioral assessments against production systems. It’s coming soon; the waitlist is open now on the certification page.
Verifiable, not decorative
A credential is only as good as its verification story, and most online certificates fail here — a PDF, a badge image, nothing anyone can actually check.
Every SAPIEN certificate lives at a permanent public URL: sapienframework.org/verify/<id>. Anyone — an employer, a client, a compliance auditor — can open that link and confirm the certificate is real, who holds it, and what level it represents. And it adds to a LinkedIn profile in one click.
This matters most for the people we expect to use the credential professionally: MSPs and consultants who need to demonstrate competence to clients, compliance teams who need documented qualifications for AI governance work, and red-teamers who want their behavioral testing skills to be legible outside their own organization. A shared, verifiable baseline is what turns “I know about AI safety” into something a third party can rely on.
The founding cohort
Practitioner certification is free right now for the founding cohort. When the cohort closes, it goes to $249.
Founding-cohort members also get a permanent cohort number on their certificate, assigned by issuance order. It stays on the certificate permanently — a record that you were doing this work before it was standard practice. We think behavioral safety assessment will be a normal part of AI deployment reviews within a few years. The people certified now will have been early, and their certificates will say so.
How to start
Go to sapienframework.org/certify and start with Awareness. It’s free, it takes under an hour, and it doesn’t expire. If it’s useful — and if you work with AI professionally, it will be — continue to Practitioner while the founding cohort is open and the fee is waived. If you’re aiming at full assessment work, join the Assessor waitlist while you’re there.
The research told us models drift under pressure. The tooling made it measurable. The missing piece has been people — trained, tested, and verifiable. That’s what this program is for.
Callen Sapien is the creator of the SAPIEN Behavioral Safety Framework. He leads behavioral AI safety research at SAPIEN Labs LLC.