data.js, so only the numbers printed in your original page are real here. On your site this file reads assets/data.js and the Sample tags disappear.What AI governance jobs actually require in 2026
Every week we read the open U.S. AI governance market one employer at a time and verify each posting on the employer's own careers page. This edition covers postings from employers. No estimates, no scraped aggregates, no third-party study.
Ten findings
If you read nothing else, read these. Every number is counted from the postings themselves.
Do AI governance jobs require a certification?
Almost never. of postings asked for no certification at all, and not one required the IAPP AIGP or any other AI-specific credential. The ones that did require something asked for a professional license, such as bar admission, a CPA, a FINRA registration or a medical license.
Each square is one posting
Every one is a license to practice a regulated profession. None is an AI certificate.
What the required credentials actually were
The preferred list is led by security and audit
Security and audit credentials total 59 mentions against 10 for the AI-specific one, roughly six to one, inside AI governance roles.
Why this happens
These employers are not staffing a new academic discipline. They are staffing a control function: testing, evidence, documentation, sign-off. The people who have demonstrably run a control function before are security people and audit people. A certificate is a claim about what you know; these roles are built around what you can be held answerable for.
What these roles pay
of the postings published a real annual range in the posting itself. The median midpoint is $205,250. The spread is wide because seniority in this field runs from analyst to executive director.
Where the middle of the market sits
Roles that require coding do not pay more
Postings needing technical AI understanding but no coding have the highest median of the four bands. "Programming preferred" has the lowest; that phrase tends to mark a more junior role rather than a more technical one.
Median pay by career pathway
The most technical pathway, model risk and data science, sits lowest: $54,000 below project and program management.
Seniority moves pay far more reliably than technical depth does
View as table
Read this carefully
Coding is not the premium, but that does not make this a non-technical field. 152 of 173 postings (88%) require or prefer prior AI or machine-learning experience. Technical literacy is table stakes. It is what gets you considered; it is not what gets you paid.
How technical are these roles really?
Most career advice collapses three different things into one. The postings separate them cleanly.
Rare, near-universal, and the minority
Adjacent engineering roles stay in the index but are flagged: they build or secure AI products rather than govern them.
Technical intensity
Where the jobs actually are
Not at the AI companies. More than a quarter of this market is financial services and insurance. Frontier AI and big tech together account for 20 of 173 postings.
Postings by sector
Fully remote is the smallest named arrangement
36% of postings do not state an arrangement at all. Of those that do, fully remote is the minority at 23 postings.
Most-named locations
Which background each role is written for
Three frameworks carry almost the whole field
Why the map looks like this
AI governance is not being invented from nothing. It is being grafted onto risk functions that already exist. A bank already has model risk management, internal audit, a second line of defense and a committee that approves things in writing. So the roles are senior because the person must be credible to that committee, mostly hybrid because the committee is somewhere, and they name frameworks a regulator will recognize. Governance follows regulation.
How senior is this field?
Very. The median posting that states a minimum wants 7 years, and exactly one posting in the sample was entry level.
Postings by seniority
How many postings would a candidate clear?
Ten or more years: 45 postings (31%). Bachelor's degree counts cover all 173 postings.
What this does and does not mean
The field is not closed. It is not hiring beginners; it is hiring experienced professionals into a new specialty. The seniority is structural: these roles have to be credible to a risk committee that already exists, and you cannot be the newest person in the building and also the person that committee believes.
Download the full reports
Three reports built from the same verified dataset. Download the PDF or read it in the browser. No email, no form.
How to read these numbers
Method
- Collection. Seven sector-specific research passes each week across employer career sites, Workday, Greenhouse, Ashby, SmartRecruiters, USAJOBS and association job boards. Target 50 to 100 new U.S. postings per week, maximum three per employer.
- Verification. Every posting is checked on the employer's own site or applicant tracking system on the verification date. Postings verified only on a third party are labeled as such and are a small minority.
- De-duplication. Each week excludes postings already published in earlier weeks, so the index describes newly found roles rather than the total stock of open roles.
- Pay. Only employer-published annual ranges are counted, annualized where stated hourly. Figures are the midpoint of each range. One lump-sum fellowship stipend is excluded.
- Skills and frameworks. Counts are keyword matches against requirements extracted from each job description. "Not specified" is reported as not specified. It is never read as "not required."
Limits worth stating
- This is a convenience sample of 173 postings collected over two weeks. It is a reading of the live market, not an estimate of the whole U.S. AI governance labor market.
- Salary figures skew toward pay-transparency states (California, Colorado, New York, Washington, Illinois), because those employers are legally obliged to publish a range.
- Some pathway and seniority groups are small, a dozen postings in places, and a single unusual posting moves a median that size. Group sizes are printed beside every figure.
- No trend claims. Week-to-week differences reflect which employers were sampled, not movement in the market. A month of pulls is the smallest honest trend unit, and this edition makes no trend claims.
Get each edition when it publishes
The index refreshes weekly. Subscribers get the edition summary, the three reports, and the postings worth knowing about before they close.
Thinking about moving into AI governance? Join a free info session.