Skip to main content
v2026.11,986 entries · CC-BY 4.0

AIRF (CASRAI)

AI Governance Course for Research: AI in Research Governance — Foundation

Examined AI governance course for research staff: AI authorship and disclosure, peer review, funding applications, integrity risk and data governance. No prerequisites.

LevelFoundationAIRAIRF (CASRAI)
3 h
Nominal
Self-paced study
9
Lessons
Self-paced
7
Domains
Exam blueprint
50
Exam items
Per attempt
70%
Pass mark
Fixed, uncurved
90
Retake wait
Days
3 yr
Valid for
From the pass

What you earn

The credential itself

Pass the examination and CASRAI issues this certificate and the post-nominal AIRF (CASRAI), signed by the Programme Director, valid for 3 years, and checkable by anyone from the verification code on its face.

Specimen. Issued as a signed PDF and as a machine-readable Open Badge; the holder and code shown are illustrative.
How it appears on your CV

Certifications

AIRF (CASRAI) — AI in Research Governance, Foundation

CASRAI·Issued Mar 2026·Credential ID CASRAI-AIR-2026-HGIVAX

casrai.org/verify/CASRAI-AIR-2026-HGIVAX

Every certificate is issued under the signature of Dr. Diana Nieves Castro, MD, Programme Director of CASRAI Certification, who sets the syllabus, blueprint and pass mark for this credential.

It certifies that its holder passed the CASRAI examination for this course on the date shown, scored against the blueprint and pass mark published on this page — and anyone can confirm it from the verification code, without an account and without contacting us.

About this credential

What it covers

Who this course is for

AI in Research Governance — Foundation (AIRF (CASRAI)) is for people who have to apply an AI policy, advise a researcher or make a first judgement when AI comes up in research work. That includes:

  • research integrity, research ethics and compliance staff;
  • journal editorial office staff and managing editors;
  • research administrators and sponsored programs staff;
  • research librarians, data stewards and repository managers;
  • graduate school staff and doctoral supervisors;
  • research IT and information governance staff;
  • researchers who want a clear working knowledge of the rules instead of half-remembered publisher policies.

It is not a machine learning course. It teaches nothing about building, training or evaluating AI models.

What you will be able to do

  • Explain in plain terms what generative AI systems do, why they produce fluent but false output such as invented references, and which failure modes matter for a given research use.
  • Explain why an AI system cannot be an author, map AI assistance to contributor roles, and decide whether the human contributors meet authorship criteria.
  • Decide whether a use of AI needs to be disclosed and where the disclosure belongs, judge whether a disclosure statement is adequate, and rewrite one that is not.
  • Recognise when a reviewer’s use of AI breaches peer-review confidentiality, and respond to a suspected AI-generated review without making an unfounded allegation.
  • Advise an applicant on using AI to prepare a funding application, including who is responsible for what is submitted under the institution’s name.
  • Classify an AI-related integrity concern against the established definitions of fabrication, falsification and plagiarism, verify a suspect reference list, and explain why an AI-detector score is a reason to look and never a finding.
  • Decide whether the basis on which research data is held allows it to be entered into a third-party AI service, list the vendor terms that need checking, and specify what must be recorded for an AI-assisted method to be evaluated.
  • Route an AI governance question to the part of the institution that owns it, and explain how narrow the research exclusions in the EU AI Act are.

The modules

  1. What these systems do, and what they get wrong. Types of AI system, why fluent falsehoods happen, and the failure modes that matter in research.
  2. Authorship and accountability. Why authorship requires accountability, how the Contributor Roles Taxonomy (CRediT) frames AI assistance, and what this means for the human authors.
  3. Disclosure: what, where and how much. Why the place of a disclosure depends on what the AI did, what does not need disclosing, and the hard cases.
  4. AI in peer review. Confidentiality, what reviewers may and may not do, the limits of AI detection, and how to handle a suspect review.
  5. AI in funding applications. Applicant responsibility, the institutional signature, the kinds of use funders treat differently, and how to advise when funders disagree.
  6. When AI meets research integrity. Fabricated references, paper mills, image integrity, honest error, and keeping screening, inquiry and finding separate.
  7. Data, consent and the tool you are about to use. Whether consent and agreements allow the use, data categories, vendor terms, and the intellectual property questions to recognise and refer.
  8. Documenting AI-assisted work. The minimum record for an AI-assisted method, model cards, datasheets for datasets, and reporting guidelines.
  9. Where the questions go. Who in an institution owns which AI question, and an orientation to the EU AI Act and its research exclusions.

Each module is written for self-paced study and includes exercises with model answers.

The examination

  • 50 scored multiple-choice questions per attempt, drawn from a larger question bank in proportion to the weight of each of the seven exam domains.
  • Pass mark: 70%.
  • You must finish the exam within 4 hours of starting it.
  • You must complete all lessons before you can start the exam.
  • If you do not pass, you can retake the exam after 90 days.

Eligibility

The Foundation tier has no prerequisites. Anyone can enroll.

The Professional tier (AIRP (CASRAI)) requires one of four routes: two years of research governance experience within the last five years, at least six months of it involving AI questions; a relevant postgraduate qualification plus one year of that experience; this Foundation credential plus one year of that experience; or an assessed adjacent credential plus one year of that experience. Professional buyers confirm at checkout that they meet one of these routes. No documents are checked.

What the credential is, and what it is not

AIRF (CASRAI) is an examined credential issued by CASRAI. It shows that you passed an examination on the governance of AI in research, and it can be verified.

  • It is not accredited by any accreditation body.
  • It carries no continuing education credit (CE, CME or CEU).
  • It does not meet any training requirement set by law, a funder or an institution.
  • It is not endorsed by any regulator, funder, publisher or professional body.
  • It certifies governance knowledge, not technical competence. It does not qualify you to evaluate an AI system’s performance, safety or bias.
  • It is not a legal or data protection qualification, and it does not certify that you or your institution comply with the EU AI Act or any other law.
  • It is not equivalent to, and not a substitute for, any general AI governance credential, such as the IAPP’s AIGP.

Validity and renewal

The certificate is valid for 3 years from the date you pass. To renew it, you pass the exam that is current at the time. Because policies on AI in research change often, the certificate attests to what you knew on the date you passed.

Price

US$119, paid once. Your access to the course does not expire.

Body of knowledge

Exam blueprint

The exam is assembled to these weights on every attempt. They are published before purchase because a blueprint you cannot see is not a blueprint.

Domains and weights7 domains · 50 scored items
DomainWeight
AI systems in research: capabilities, limits and failure modesD1
18%9 items
Authorship, contributorship and disclosure of AI assistanceD2
20%10 items
AI in peer review and research assessmentD3
12%6 items
AI in funding applications and funder policyD4
10%5 items
Research integrity risk: fabrication, paper mills and image integrityD5
16%8 items
Data governance, consent, documentation and intellectual propertyD6
14%7 items
Institutional governance and the regulatory overlayD7
10%5 items
Total100%

Every attempt has this many items from each domain. The items themselves are drawn at random from the bank within those fixed per-domain quotas, so no two attempts are the same paper.

Syllabus

Syllabus: 9 lessons, 3 hours

Every lesson and every learning outcome is listed. The lesson bodies open on enrolment, and the exam opens once every lesson is marked complete.

9 lessons · 3 h

Lesson material opens once you are enrolled. The full syllabus and every learning outcome are shown here so you can judge the coverage before you pay.

  1. Lesson 1

    Module 1

    1 lesson · 15 min

    1. Locked.What AI systems do, and what they get wrong15 min · Locked

      On completion you will be able to

      • Classify a described system as predictive, generative, multimodal, retrieval-augmented or agentic, and state one governance consequence of the classification.
      • Explain in non-technical language why a language model can produce a fabricated citation that looks correct.
      • Identify the failure modes most relevant to a described research use of AI.
      • Explain why version instability makes an undocumented AI-assisted method unreproducible.
      • Recognise when a question requires technical expertise outside governance competence and name who should answer it.
  2. Lesson 2

    Module 2

    1 lesson · 15 min

    1. Locked.Authorship and the accountability argument15 min · Locked

      On completion you will be able to

      • Construct, in your own words, the argument for why an AI system cannot be an author.
      • Map a described AI-assisted contribution onto CRediT roles and identify the human who directed and verified it.
      • Determine whether each named human contributor in a described case meets the ICMJE authorship criteria.
      • Explain why the AI-authorship question and the gift-authorship question are the same question.
  3. Lesson 3

    Module 3

    1 lesson · 15 min

    1. Locked.Disclosure: what, where and how much15 min · Locked

      On completion you will be able to

      • Determine whether a described use of AI requires disclosure.
      • Decide whether a required disclosure belongs in the acknowledgments or the methods, and justify the choice by what the AI did.
      • Assess a disclosure statement for adequacy against a supplied journal policy.
      • Rewrite an inadequate disclosure statement so that a reader could evaluate or reproduce the work.
      • Explain why AI output cannot be cited as a source.
  4. Lesson 4

    Module 4

    1 lesson · 15 min

    1. Locked.AI in peer review15 min · Locked

      On completion you will be able to

      • Explain why entering a manuscript or proposal under review into a third-party AI service breaches confidentiality at the moment of upload.
      • Distinguish prohibited reviewer uses of AI from uses that raise only a disclosure or policy question.
      • Identify the indicators that legitimately raise a question about an AI-generated review.
      • Draft a proportionate inquiry to a reviewer that does not allege misconduct.
      • Explain why an AI-detector score cannot support a finding against an individual.
  5. Lesson 5

    Module 6

    1 lesson · 30 min

    1. Locked.When AI meets research integrity30 min · Locked

      On completion you will be able to

      • Classify a described AI-related concern as possible fabrication, falsification, plagiarism, questionable practice or honest error, using the established definitions.
      • Verify a reference list and distinguish fabricated references from mis-transcribed ones.
      • List the submission indicators that suggest paper-mill or mass-generated content.
      • Explain why an image concern is referred for specialist assessment and raw-data request.
      • Separate the screening, inquiry and finding stages in a described case and name the next correct step.
  6. Lesson 6

    Module 7

    1 lesson · 15 min

    1. Locked.Data, consent and the tool you are about to use15 min · Locked

      On completion you will be able to

      • Determine whether a proposed use of a third-party AI service is compatible with the consent, agreement or approval under which the data is held.
      • Identify the consent, confidentiality and data protection questions a described use raises, and when to refer them.
      • List the vendor terms that must be checked before research data enters an AI service, and explain why each matters.
      • Explain why consumer-tier use can defeat an institutional enterprise agreement.
      • Describe the intellectual property questions raised by AI-generated research outputs without stating a legal conclusion.
  7. Lesson 7

    Module 8

    1 lesson · 15 min

    1. Locked.Documenting AI-assisted work15 min · Locked

      On completion you will be able to

      • Specify the elements that must be recorded for an AI-assisted method to be evaluated or reproduced.
      • Write a methods paragraph for a described AI-assisted analysis that a replicator could follow.
      • Identify what is missing from a supplied methods description of AI use.
      • Describe the questions model cards and datasheets for datasets ask, and why they are useful.
  8. Lesson 8

    Module 9

    1 lesson · 30 min

    1. Locked.Where the questions go30 min · Locked

      On completion you will be able to

      • Identify the institutional owner of a described AI governance question.
      • Recognise an AI governance question that has no clear owner and propose how to route it.
      • Apply supplied text of Articles 2(6) and 2(8) to decide whether a described university AI use falls within the research exclusions.
      • Explain why the EU AI Act research exclusions are narrower than institutions often assume.
      • Explain how existing research integrity principles answer most AI governance questions.
  9. Lesson 9

    Module 5

    1 lesson · 30 min

    1. Locked.AI in funding applications30 min · Locked

      On completion you will be able to

      • Advise an applicant on which uses of AI in preparing an application carry the most risk, and why.
      • Apply a supplied funder policy extract to decide whether a described use should be disclosed and where.
      • Explain why AI-generated inaccuracies in a proposal are an institutional matter as well as a personal one.
      • Classify fabricated preliminary data or citations in an application under existing misconduct definitions.
      • Draft institutional advice for a researcher facing two funders with different AI rules.

Assessment

How the exam works

The exam is closed-book and multiple choice. Each attempt draws a fresh form to the domain weights above, so no two attempts are the same paper and no answer key circulates. The pass mark is 70% and is fixed — there is no curve, no quota and no adjustment by cohort.

Scoring is immediate. Pass, and every item is reviewed with the option you chose, the correct option and its rationale. If you do not pass, you see your overall score and your score per domain (smaller domains are grouped together); which items you missed, and their answers, are held back so your resit is still a real assessment. A failed attempt may be retaken after 90 days. The wait exists so a retake is a second attempt at the material rather than a second attempt at remembering the paper.

The exam opens once you have marked every lesson of the course complete. Forms are assembled to the blueprint above or not at all: where a domain cannot yet be sampled to that standard, the exam declines to start rather than issue an unbalanced paper. Your material and progress are never affected, and there is no time limit on when you sit.

Passing candidates are issued a certificate with a verification code, valid for 3 years from the date of the pass. Anyone can check that code on our public verification page without an account and without contacting us. To renew, pass the current examination again: renewal opens 180 days before the certificate expires, and a pass issues a renewed certificate valid for a further 3 years from that date.

Certifying authority

Signed by Dr. Diana Nieves Castro, MD

CASRAI’s certification programme is academically and medically directed by Dr. Diana Nieves Castro, MD, Programme Director, who sets the syllabus and blueprint for this credential and signs every certificate issued under it.

CASRAI has maintained the terminology and reporting standards of research administration for over a decade. It assumed leadership of the CRediT contributor-roles taxonomy in 2014 and carried it through to adoption as ANSI/NISO Z39.104-2022, now in use by publishers, funders and institutions worldwide. This examination is drawn from that body of work.

What the credential certifies: that its holder passed the CASRAI examination for this course on the date shown, scored against the blueprint and pass mark published above. It is valid for 3 years, and anyone can check it from the verification code on its face.

The ladder

Other levels of AI in Research Governance

Levels are independent purchases. There is no prerequisite chain — sit whichever level matches the work you already do.

AIRP (CASRAI)Professional

AI in Research Governance — Professional

Examined credential for people who write and apply research AI policy: authorship, peer review, funder rules, misconduct cases, data and vendor…

Nominal
5 h
Lessons
12
Domains
7
Exam items
75
Cut score
70%

All CASRAI credentials

Common questions

AI Governance Course for Research: AI in Research Governance — Foundation: common questions

Is there any certification for AI governance?

Yes. General credentials such as the IAPP's AIGP cover AI governance across industries, while AI in Research Governance — Foundation is an examined credential issued by CASRAI that focuses on research: authorship and disclosure of AI assistance, AI in peer review and funding applications, research integrity risk, and research data governance. It certifies governance knowledge, not the ability to build or evaluate AI systems, and it is not a substitute for a general AI governance credential.

Is there a free course on AI governance?

This course is not free. It costs US$119, paid once by card through Stripe checkout, and your access to the course does not expire. The price covers nine self-paced lessons (3 hours of study) and the online exam that leads to the certificate.

How do I get my AI certification?

The Foundation tier has no experience prerequisite, so you can enroll and start right away. Complete all nine lessons to unlock the exam: 50 multiple-choice questions with four options each, which you must finish within 4 hours of starting, and a pass mark of 70%. If you do not pass, you can retake the exam after a 90-day waiting period.

Can AI be listed as an author on a research paper?

No, under the ICMJE Recommendations. ICMJE says chatbots such as ChatGPT and other AI-assisted tools should not be listed as authors, because they cannot be responsible for the accuracy, integrity and originality of the work. Authorship and disclosure of AI assistance is the most heavily weighted exam domain in this course, at 20%.

Can AI help in writing a research paper?

Yes, if the use is disclosed and the human authors take responsibility for the result. ICMJE asks authors to describe how they used AI in the cover letter and in the manuscript, to review and edit AI-generated content carefully, and to be able to confirm that the paper contains no plagiarism, including in AI-produced text and images. Module 3 of the course covers where a disclosure belongs and how to judge whether a disclosure statement is adequate.

How long is the AI governance certificate valid?

The certificate is valid for 3 years. You renew it by passing the current exam again, and the renewal window opens 180 days before expiry. Each certificate has a public verification page at casrai.org/verify/ followed by the certificate code.

Referenced across the research world

University of Cambridge logoColumbia University logoCrossref logoUniversity of Edinburgh logoHarvard University logoUniversity of Oxford logoPrinceton University logoStanford School of Medicine logoUniversity College London logoORCID logoUniversity of Cambridge logoColumbia University logoCrossref logoUniversity of Edinburgh logoHarvard University logoUniversity of Oxford logoPrinceton University logoStanford School of Medicine logoUniversity College London logoORCID logo
  • University of Cambridge logo
  • Columbia University logo
  • Crossref logo
  • University of Edinburgh logo
  • Harvard University logo
  • University of Oxford logo
  • Princeton University logo
  • Stanford School of Medicine logo
  • University College London logo
  • ORCID logo

View CASRAI adoption →