Skip to content

RA · Foundations

AI and machine learning for risk analysis: what to learn, in what order, and why

A learning guide to AI and machine learning for risk analysis: core ideas, an order of study, key frameworks and data, open courses and how expertise is shown.

As of
17 November 2026
చదవడం
12 minutes
ఆధారాలు
11 sources, cited in order
స్థాయి
Introductory to intermediate
Study time
15 minutes
NEXUS TWIN · WF-1 · SURFACESEED 3cf66c5cRA-F04.DISCIPLINE.AI-MACHINE-LEARNING-RISKxyzSCHEMATIC · SURFACEST-01 · 0.347 0.581ST-02 · 0.180 0.167
WF-1 · surface Schematic wireframe twin (surface, learning and optimisation). An illustration, not data.

What you will be able to do

  • Explain data leakage and why test data must stay apart.
  • Name the four functions of the NIST AI Risk Management Framework.
  • Distinguish machine learning as a tool for risk work from AI as a source of risk.
  • Write a short model note that states purpose, data, tests and limits.
స్థాయి
Introductory to intermediate
Study time
15 minutes
Before you start
The learning guide to the Faculty of Risk Science and Methods, or equivalent background
Framework
4 functions
in the NIST AI Risk Management Framework: govern, map, measure and manage2
Principles
May 2019
when the OECD AI Principles were adopted6
Forecasting
1,380
verification targets in the GraphCast evaluation4

01What AI and machine learning for risk analysis covers

What AI and machine learning for risk analysis covers

AI and machine learning for risk analysis is the Academy discipline of AI and machine learning for risk analysis, one of the disciplines of the Faculty of Risk Science and Methods. It covers how software learns patterns from data, how those patterns are tested before anyone relies on them, and how the resulting systems are governed.

The field has two sides that a learner needs to keep apart. In the first, machine learning is a method for risk work: forecasting weather, flagging anomalies, mapping exposure from imagery. In the second, an AI system is itself a source of risk, to individuals, to organisations and to society. The NIST AI Risk Management Framework is written for the second side. NIST describes it as intended for voluntary use, to improve the ability to build trustworthiness considerations into the design, development, use and evaluation of AI products, services and systems 1.

The discipline sits beside model risk and validation, geospatial analysis, Earth observation and GeoAI and risk indicators and composite indices. It relies on the statistics and probability taught in the Faculty of Risk Science and Methods as a whole.

02Why this field matters for risk work

Why this field matters for risk work

Machine learning has moved from research into operations. The European Centre for Medium-Range Weather Forecasts (ECMWF) built a data-driven forecast model, the Artificial Intelligence/Integrated Forecasting System (AIFS), to explore the machine learning technologies behind recent models. It publishes AIFS products on its OpenCharts service alongside charts of its physics-based system 3. A learner who works with forecasts will meet such models.

The research behind this shift is open to read. The GraphCast paper reports a model trained directly from reanalysis data. It predicts hundreds of weather variables over ten days at 0.25 degree resolution in under one minute. The authors report that it outperforms the most accurate operational deterministic system on 90 per cent of 1,380 verification targets 4. Those are the authors' reported results, and a learner should read them as such: a claim about one evaluation, not a general ranking.

The same tools raise governance questions. The OECD AI Principles were adopted in May 2019. They promote AI that is innovative and trustworthy and that respects human rights and democratic values 6. The OECD notes that governments must track and understand AI incidents and hazards, and runs a monitor for them 6. The Financial Stability Board published a report in November 2017 on the financial stability implications of the growing use of AI and machine learning in financial services 7. A risk professional is now expected to read both the models and the rules around them.

03The main parts of AI risk work

The main parts of AI risk work

A useful map separates four kinds of work. Data work covers where training data come from, what they leave out and how they are split for testing. Model work covers how a method is chosen, trained and compared with a simpler baseline. Evaluation work covers whether results hold on new data, in new places and under stress. Governance work covers who is accountable, how incidents are recorded and which rules apply.

The NIST AI RMF gives governance work a shape. Its core is composed of four functions: govern, map, measure and manage, and the framework document, NIST AI 100-1, was published in January 2023 2. Govern is a cross-cutting function: the framework describes the culture and processes set up under it as informing the other three 2. Figure 2 sets out the four parts of the work.

The European Union takes a legal route. The European Commission describes the AI Act, Regulation (EU) 2024/1689, as setting risk-based rules for developers and deployers of AI for specific uses, with categories that include unacceptable risk, high risk and minimal or no risk 5. A learner does not need to know every article. A learner does need to know that the two routes, a voluntary framework and a binding regulation, ask related questions about the same systems.

Fig. 02Four kinds of work in AI risk analysis

Four kinds of work in AI risk analysisFour stacked layers naming the kinds of work in AI risk analysis: data work, model work, evaluation work and governance work, with governance highlighted as the layer that informs the others.01Data workSources, gaps, splits02Model workMethod, training, baselines03Evaluation workNew data, new places, stress04Governance workAccountability, incidents, rules
A teaching map of the work, with governance drawn from the NIST framework 2. A weak layer weakens the whole.2Sources: NIST AI 100-1; Academy desk teaching map. Figure CC BY 4.0. As of 17 November 2026.
Figure 2 in words

Four layers stacked from top to bottom. Data work: sources, gaps and splits. Model work: method, training and baselines. Evaluation work: new data, new places and stress. Governance work, highlighted: accountability, incident records and applicable rules, which the NIST framework describes as informing the other functions.

04A suggested learning order in four stages

A suggested learning order in four stages

The Academy describes study at four levels: introductory, intermediate, advanced and integrative. The four stages below follow the same arc for AI and machine learning for risk analysis. They are a suggestion, and a learner can enter at any stage.

Stage one: data and evaluation. Start with the data. Learn what a training set, a validation set and a test set are for, and what data leakage is. The scikit-learn documentation defines data leakage as information that would not be available at prediction time being used when the model is built, and gives guidance on avoiding it 8. This stage also covers basic statistics and the habit of comparing any model with a simple baseline.

Stage two: models and their limits. Next, learn the main families of method and what each can and cannot do. An introduction such as Elements of AI is a free way in for people who want to know what AI is, what can and cannot be done with it, and how to start creating methods 9. A university course such as MIT 6.036 Introduction to Machine Learning goes further into the methods 10. Read also the argument that, for high-stakes decisions, inherently interpretable models should be preferred to explanations of black box models 11.

Stage three: governance frameworks. Then study the frameworks. Read the NIST AI RMF and its core functions 1 2. Read the OECD AI Principles 6 and the European Commission summary of the AI Act 5. Compare what each asks of an organisation that builds or uses a model.

Stage four: use in operational risk work. Finally, apply the ideas. Read how a forecasting centre introduces a data-driven model, with its stated training data and its stated limits 3, and how researchers evaluate one 4. Practise writing a short note that says what a model is for, how it was tested, where it fails and who reviews it.

The order is a suggestion. A practitioner may begin at stage three and return to stage one later. What matters is that each new idea has the ideas beneath it in place.

Fig. 01Four stages of study in AI and machine learning for risk analysis

Four stages of study in AI and machine learning for risk analysisA rising ladder of four steps: data and evaluation, models and their limits, governance frameworks, and use in operational risk work, each with the references a learner meets at that step.01Data and evaluationSplits, leakage, baselines02Models and limitsMethods, interpretable models03Governance frameworksNIST, OECD, EU AI Act04Use in risk workForecasting, model notes
The suggested learning order of this guide, with the references that anchor each stage 8,9,10,11,1,2,6,5,3,4. The order is a suggestion, not a required curriculum.8,9,10,11,1,2,6,5,3,4Sources: scikit-learn; Elements of AI; MIT OpenCourseWare; NIST; OECD; European Commission; ECMWF. Figure CC BY 4.0. As of 17 November 2026.
Figure 1 in words

Four steps rise from left to right. Step one, data and evaluation: training, validation and test data, and data leakage. Step two, models and their limits: methods, introductory courses and interpretable models. Step three, governance frameworks: the NIST AI Risk Management Framework, the OECD AI Principles and the EU AI Act. Step four, highlighted, use in operational risk work: data-driven forecasting and the model note.

Fig. 03The four core functions of the NIST AI RMF

The four core functions of the NIST AI RMFA four-stage cascade of the functions in the NIST AI Risk Management Framework core: govern, map, measure and manage, with govern described as the cross-cutting culture and processes that inform the rest.GOVERNCulture and rolesCross-cuttingMAPContextIntended useMEASUREAnalyseTrack riskMANAGEPrioritiseAct on risk
The four functions of the framework core as named in NIST AI 100-1 2.2Sources: NIST AI 100-1 (January 2023). Figure CC BY 4.0. As of 17 November 2026.
Figure 3 in words

Four stages run from left to right: govern, map, measure and manage. Govern holds the culture, roles and processes of AI risk management. Map establishes context. Measure analyses and tracks risk. Manage acts on the risks that have been measured. The framework document states that governance is cross-cutting and informs the other functions.

05Six core ideas every learner should hold

Six core ideas every learner should hold

Six ideas carry most of the weight in this discipline. Each is simple to state and takes practice to use.

Evaluation comes before trust. A model is only as useful as its test. A result on the data used to build it says little about new cases 8.

Data decide a great deal. The AIFS is trained on a subset of the ERA5 reanalysis for 1979–2018 and fine-tuned on operational forecast data from 2019 to 2020, and its authors state that at one degree resolution it is still behind the operational system for surface temperature 3. What a model learned from sets what it can do.

Compare against the incumbent. GraphCast was evaluated against the most accurate operational deterministic system, not in isolation 4. A new model earns its place by beating what is already used, on the same test.

Interpretability is a design choice. For high-stakes decisions there is a serious argument for models that are interpretable from the start 11.

Risk management covers the whole life of a system. The NIST framework is about design, development, use and evaluation, not only a final check 1.

Rules differ by place and are changing. NIST notes that AI RMF 1.0 is being revised 1. The European Commission page states that the AI Act entered into force on 1 August 2024 and became applicable on 2 August 2026, with some exceptions, and that rules for certain high-risk areas apply from 2 December 2027 5. Always read the current text.

06Canonical references and data to know

Canonical references and data to know

A learner should know a short list of references and what kind of evidence each holds. Primary documents come first; summaries help only after them.

For AI risk management: the NIST AI RMF page 1 and the framework document NIST AI 100-1 2. For law and principles: the EU AI Act summary 5 and the OECD AI Principles 6. For the financial sector: the FSB report on AI and machine learning in financial services 7. For forecasting practice: the ECMWF AIFS newsletter article 3 and the GraphCast paper 4. For method: the argument on interpretable models 11 and the scikit-learn guidance on common pitfalls 8. Datasets are not listed here because the right dataset depends on the problem; learners should always record the source, date and licence of any data they use.

07Skills that make the knowledge usable

Skills that make the knowledge usable

Knowledge becomes useful through a small set of skills. The discipline treats these as its practical core.

Splitting data honestly. Keep test data apart and avoid leakage. Comparing with a baseline. Report how a model does against a simple alternative. Reading an evaluation. Ask what was measured, over which cases, and what the model was not tested on. Documenting a model. Say what it is for, what data it used and what it must not be used for. Recording incidents. Keep a log of failures and near misses, as the OECD monitor does for AI incidents at global level 6. Explaining limits. Write plainly for a decision maker who is not a specialist.

08Open courses and pathways

Open courses and pathways

Several free courses are open to the public and were opened at the provider on 17 November 2026. Elements of AI is a series of free online courses created by MinnaLearn and the University of Helsinki. Its introductory course needs no complicated mathematics or programming, and its Building AI course recommends some basic Python skills 9.

MIT 6.036 Introduction to Machine Learning was taught in Fall 2020 as an undergraduate course in electrical engineering and computer science. It is part of the Open Learning Library, which is free to use, and a learner can view the materials without enrolling 10. The page presents course materials; it does not say that completing them leads to a qualification, and this guide does not either.

A suggested pathway is the introduction first, then the university course, then the frameworks of stage three. Where a course page lists no assessment, treat it as study material rather than a record of skill.

09How expertise is shown

How expertise is shown

Expertise in this field is shown by work that others can check. The Academy records learning, not credentials.

Finishing a course gives a completion record. It is not a certificate, an accreditation or a qualification, and it does not state that a person is competent in a profession. Achievements on the Academy recognise assessed learning only.

The Academy catalogue lists two courses for this discipline: AI and machine learning for risk analysis: foundations and practice, at the intermediate level, and AI and machine learning for risk analysis: frontier methods, at the advanced level. On 17 November 2026 the catalogue page shows the first as planned and the second as planned. Nothing is open for enrolment until it is published in the NXS Learning catalogue, so this guide does not point to an Academy course as a place to enrol.

Outside the Academy, the strongest evidence of expertise is a body of work: a documented model note that cites its data, tests and limits, a reproducible evaluation that a reviewer can rerun, or a short brief that compares a model with a baseline. Learners are encouraged to keep such a record from their first course.

10Emerging areas to watch

Emerging areas to watch

Several areas are changing what a risk analyst needs to know. Each is a place to read on once the core is in place.

Data-driven forecasting. ECMWF states that an important next step for the AIFS is to extend it to ensemble forecasts 3. Sector profiles. NIST reports a concept note, released on 7 April 2026, for an AI RMF profile on trustworthy AI in critical infrastructure 1. Incident monitoring. The OECD runs an AI Incidents and Hazards Monitor 6. Revision of frameworks. NIST states that AI RMF 1.0 is being revised 1.

Related work across the Nexus network gives wider context. See responsible technology research at the Global Centre for Risk and Innovation, AI and model risk in finance at the Global Risks Alliance, digital and AI governance at the Global Risks Forum and the proposed Nexus Standards on AI and models. These pages describe programmes and proposals. A link to them implies no endorsement of the sources this guide cites.

What this guide does not claim: it is suggested learning, not a curriculum that anyone must follow, and it is not professional, legal, medical or technical advice. It names programmes, courses and bodies as places to learn and implies no partnership with any of them. It reports figures as their issuers published them and draws no new estimates from them. Risk Academy does not certify competence, and a completion record is not a professional qualification. The Academy records no partners, cohorts or learner outcomes for this faculty, and this guide implies none. The Nexus Standards are proposed and in draft; nothing here treats them as adopted.

ప్రశ్నలు

Questions and answers

Where should a complete beginner start in AI and machine learning for risk analysis?

Start with a free introductory course such as Elements of AI, then stage one of this guide on data and evaluation.

Do I need to program to follow this guide?

Not for stages one and three. Some basic Python helps for building models, and the Building AI course of Elements of AI recommends it.

Is the NIST AI Risk Management Framework mandatory?

NIST describes it as intended for voluntary use. Other rules, such as the EU AI Act, are legal texts with their own scope.

What is data leakage?

Using information in building a model that would not be available when the model makes a real prediction. It makes test results look better than real performance.

Do courses in this discipline lead to a qualification?

No. A completion record shows that you finished a course. It is not a certificate or a professional qualification.

నిబంధనలు

Glossary

Machine learning
Methods that learn patterns from data rather than following only rules written by hand.
Data leakage
Use of information in model building that would not be available at prediction time.
ప్రాథమిక స్థితి
A simple method against which a more complex model is compared.
Interpretable model
A model whose reasoning a person can follow, by design rather than by later explanation.
Reanalysis
A consistent record of past weather made by combining observations with a model.
AI incident
An event in which an AI system caused or contributed to harm or a near miss.

Check your understanding

  1. Why compare a model with a simple baseline?

    A complex model earns its place only if it does better than a simple method on the same test.

  2. What does the govern function do in the NIST framework?

    It is cross-cutting: the culture and processes set under it inform the map, measure and manage functions.

  3. Why is a test on training data not enough?

    It says little about new cases; leakage or overfitting can make results look better than real performance.

Continue with a course

మూలాలు

References

11 sources, numbered in the order they are first cited
  1. U.S. National Institute of Standards and Technology (2026). AI Risk Management Framework. www.nist.gov. Accessed 2026-11-17.
  2. U.S. National Institute of Standards and Technology (NIST AI 100-1) (2023-01). Artificial Intelligence Risk Management Framework (AI RMF 1.0), NIST AI 100-1. nvlpubs.nist.gov. Accessed 2026-11-17.
  3. European Centre for Medium-Range Weather Forecasts (2024-01). AIFS: a new ECMWF forecasting system. www.ecmwf.int. Accessed 2026-11-17.
  4. Science (Lam et al.) (2023). GraphCast: Learning skillful medium-range global weather forecasting. doi.org. Accessed 2026-11-17.
  5. European Commission (2026). AI Act: regulatory framework for AI. digital-strategy.ec.europa.eu. Accessed 2026-11-17.
  6. Organisation for Economic Co-operation and Development (OECD.AI) (2026). OECD AI Principles overview. oecd.ai. Accessed 2026-11-17.
  7. Financial Stability Board (2017-11). Artificial intelligence and machine learning in financial services. www.fsb.org. Accessed 2026-11-17.
  8. scikit-learn developers (2026). Common pitfalls and recommended practices (data leakage). scikit-learn.org. Accessed 2026-11-17.
  9. University of Helsinki and MinnaLearn (2026). Elements of AI. www.elementsofai.com. Accessed 2026-11-17.
  10. MIT OpenCourseWare (2020). Introduction to Machine Learning (6.036), Fall 2020. ocw.mit.edu. Accessed 2026-11-17.
  11. Nature Machine Intelligence (Rudin) (2019). Stop explaining black box machine learning models for high stakes decisions and use interpretable models instead. doi.org. Accessed 2026-11-17.

రికార్డు

Editorial record

Facts and course states are given as of 17 November 2026. This guide is reviewed at least every 270 days and whenever a cited report publishes a new edition or a course of the faculty opens.

Published by the Academy desk of Risk Academy. To report an error or request a correction, write through the Risk Academy contact page; corrections are dated and listed with the page. The Academy's statement on AI use explains how AI assistance is used and reviewed. Text, visuals and data are licensed CC BY 4.0, except where a cited source states other terms.

Cite this article

Academy desk, Risk Academy (2026, November 17). AI and machine learning for risk analysis: a guide. Risk Academy. https://risksacademy.com/risk-science/learning-ai-and-machine-learning-for-risk-analysis-guide/

Republish

Republish or translate this article free of charge, in print or online, under CC BY 4.0: keep the byline and Risk Academy, link to the original and note any change. Figures, maps and twins made by the institutions carry the same licence; third-party data keeps its own credit line.

Attribution to paste
<p>AI and machine learning for risk analysis: what to learn, in what order, and why — Academy desk, Risk Academy, Risk Academy. First published 17 November 2026 at <a href="https://risksacademy.com/risk-science/learning-ai-and-machine-learning-for-risk-analysis-guide/">https://risksacademy.com/risk-science/learning-ai-and-machine-learning-for-risk-analysis-guide/</a>. Republished under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/).</p>

Share

EmailLinkedInBlueskyMastodon