← All courses

Awareness training · AI governance

Safe Use of Generative AI at Work Training

Part of the ISO/IEC 42001:2023 learning path → Step 1

Online training course to use generative AI tools at work without leaking data, publishing errors or breaking client trust.

This awareness training course turns safe use of generative AI into clear working rules. It covers the risks NIST names in its Generative AI Profile, the data you must never paste, approved tools and account settings, prompt injection in connected assistants, checking outputs by type, intellectual property, disclosure and records, and what to do when something goes wrong. A business services company with centres in Manila and Mumbai runs through the course.

Secure payment by Stripe · for yourself or for your team

Add another course and save 15%, up to 25%

What you get

  • 11 video modules

    About 104 minutes, with captions. Watch at your own pace on any device.

  • Practice after every module

    Short questions check your understanding before you move on.

  • Final exam

    Multiple choice, with one retake if you need it.

  • Built by experts

    Industry experts, domain specialist trainers and globally renowned auditors with 30+ years in ISO and GRC.

A certificate you can prove, and share

Verifiable certificate with a QR code anyone can scan, plus a digital badge to add to your LinkedIn profile.

See how verification works →

Save when you take more than one course

The discount is added automatically in your basket.

  • 15% off2 courses
  • 20% off3 courses
  • 25% off4 or more

Standard and editionNIST AI 600-1; EU AI Act (Regulation (EU) 2024/1689)

Last updated9 November 2026

Preview this course

Experience the learning
before you enrol.

Watch a sample from Safe Use of Generative AI at Work Training before you enrol.

Safe Use of Generative AI at Work TrainingSample

What learners say

Rated by industry professionals like you

Every course is built by the experts behind 8,100+ client engagements in 60+ countries, so you learn what auditors and regulators actually expect. Here is what professionals said after finishing their course.

Informative

10/10 for Philippines Data Privacy Act and DPO Training

  • Built by the experts behind 8,100+ client engagements in 60+ countries
  • Certificate verifiable online
  • 12 months' access

10/10

“Informative”

Philippines Data Privacy Act and DPO Training

US$99list price US$149

Enrol now

10/10

“Engaging”

Supplier Evaluation & Second-Party Audits

US$99list price US$149

Enrol now

10/10

“Excellent”

Nonconformity Management

10/10

“Truly enjoyed”

Risk Management (ISO 31000:2018)

10/10

“Good course”

HIPAA Awareness

10/10

“Amazing course”

ISO/IEC 27001:2022 Awareness

10/10

“Very helpful”

ISO 9001:2026 Awareness

Best value

All-Access Pass

A pharma and life sciences professional from the USA started with one course, then upgraded to the All-Access Pass.

Every Sterling Academy course for 12 months.

A selection of comments from learners who finished a course and answered our feedback request, shortened from their own words. Client names available on request.

Learning outcomes

Knowledge you can put to work.

Connect the subject to the responsibilities and decisions people encounter in their roles.

  • Name the main risks of generative AI at work, as the NIST Generative AI Profile describes them
  • Decide what information may and may not be entered into a generative AI tool
  • Use only approved tools, accounts and settings, and know why they differ from personal ones
  • Recognise prompt injection and other risks of assistants connected to e-mail, files and systems
  • Check AI outputs for accuracy, sources, intellectual property and fitness for purpose before use
  • Disclose AI use where required, keep records and report mistakes and incidents quickly

Part of: AI Governance and ISO 42001 Training

Inside the course

A clear path through the essentials.

11 modules, 11 practice checkpoints and a final exam. Video runtimes are approximate; allow additional time for the practice questions and the final exam. Modules marked optional can be skipped; the others are studied in any order.

Part 1Module 1Approx. 9 min

  1. Why safe use needs rules9 min

    Explains why safe use of generative AI at work needs clear rules, the public sources the course uses, and the three moments that matter.

    What this module covers

    Opens the Safe Use of Generative AI at Work training course with a normal Tuesday at G-H-J Business Services, headquartered in Singapore with centres in Manila and Mumbai: a claim file pasted into a personal account, an unchecked reply promising a refund, and untested code. Defines generative AI as the course uses it, including assistants connected to e-mail and files. Explains Article 4 of the EU AI Act on AI literacy, as its wording was replaced in July 2026, and NIST's Generative AI Profile, NIST AI 600-1, with its twelve risks. Sets out how the course differs from the AI literacy course, what learners will be able to do, and two common myths.

    What you will be able to do

    • Describe the three moments that matter: what goes in, what comes out, what you do
    • Explain what Article 4 of the EU AI Act asks of organisations on AI literacy
    • Recognise why a confident answer is not necessarily a correct one

    Why it matters

    The tool accepts anything you give it, so clear rules for what goes in and what comes out keep everyday AI use safe.

    Practice: Practice: Why safe use needs rules · 3 questions

Part 2Module 2Approx. 10 min

  1. The risks NIST names10 min

    Walks through the twelve generative AI risks NIST names in its Generative AI Profile, and the few that office work meets every day.

    What this module covers

    Starts with a whiteboard of about thirty worries at G-H-J's head office, sorted using NIST's public list. Explains how NIST describes risks unique to generative AI or made worse by it, then takes all twelve in three groups of four, from confabulation and data privacy to information security, intellectual property and the value chain. Shows how G-H-J sorted them, which Sterling recommends as a starting point, noting that NIST does not rank them. Looks closely at confabulation, human-AI configuration and automation bias, data privacy through leakage, memorisation and inference, and environmental impacts, with a bank in Toronto that bans every tool.

    What you will be able to do

    • Name the twelve risks in NIST's Generative AI Profile
    • Identify the five risks that office work meets every day
    • Explain confabulation and automation bias in plain words

    Why it matters

    Many of the twelve risks arise from how a tool is used, so naming them shows which ones only you and your organisation can manage.

    Practice: Practice: The risks NIST names · 3 questions

Part 3Module 3Approx. 9 min

  1. Data you must never paste9 min

    Sets out the data you must never paste into a generative AI tool, a three-colour way to decide the rest, and how to cut the risk.

    What this module covers

    Opens with a claims processor at G-H-J's Manila centre who pasted a whole claim file, with medical details, into a personal chat account. Links pasting to NIST's data privacy risk, then gives the never-paste list Sterling recommends: passwords and keys, payment card numbers, sensitive personal data, and client data a contract keeps in approved systems. Explains red, amber and green, built on information classification under ISO/IEC 27001, client AI notes, and four habits for amber data: remove, replace, reduce and re-check. Covers files, screenshots, recordings and connected assistants, three questions to ask before every paste, and what is usually fine.

    What you will be able to do

    • List the information that must never go into an AI tool
    • Classify information as red, amber or green before pasting it
    • Remove identifiers and cut text to the minimum the task needs

    Why it matters

    Once data is pasted into a tool your organisation does not control, nobody can see where it goes, so the safest data never goes in.

    Practice: Practice: Data you must never paste · 3 questions

Part 4Module 4Approx. 10 min

  1. Approved tools, accounts and settings10 min

    Explains why approved generative AI tools, business accounts and the right settings matter, and what to do when a new AI feature appears.

    What this module covers

    Starts with a finance analyst in G-H-J's Mumbai team who used the approved app on her phone, but signed in with a personal account. Explains why only business accounts run on terms and controls the organisation agreed, and where approved tools come from under ISO/IEC 42001 control A.9.2 and NIST's acceptable use guidance. Shows G-H-J's approved list by type of tool, intended use under control A.9.4, four account checks Sterling recommends, and supplier checks under control A.10.3 and ISO/IEC 27001. Gives four steps for when a new AI button appears, covers contractors and temporary staff, and explains the risk of shadow AI.

    What you will be able to do

    • Use work data only in approved tools on business accounts
    • Check your account, settings, add-ons and shared links
    • Pause, ask and request before using a new AI feature for work

    Why it matters

    The same tool can be safe on one account and unsafe on another, so staying inside what was approved keeps work data under control.

    Practice: Practice: Approved tools, accounts and settings · 3 questions

Part 5Module 5Approx. 9 min

  1. Connected assistants and prompt injection9 min

    Covers AI assistants connected to e-mail and files, direct and indirect prompt injection, and the simple limits that keep them safe.

    What this module covers

    Opens with a test during G-H-J's pilot, where a hidden instruction in an e-mail told a connected assistant to forward invoices to an outside address. Explains what a connected assistant is, NIST's description of direct and indirect prompt injection, and where hidden instructions can sit. Sets out three limits Sterling recommends: least access, a person confirming before the assistant sends, pays, deletes or shares, and reporting odd behaviour. Links confirmation to human oversight in the ISO/IEC 42001 guidance, touches on data poisoning and value chain risk with a software company in Boston, and shows how G-H-J ran its eight-week pilot and the warning signs to watch for.

    What you will be able to do

    • Explain the difference between direct and indirect prompt injection
    • Apply least access and confirmation before a connected assistant acts
    • Recognise and report signs that an assistant has been manipulated

    Why it matters

    One hidden instruction can reach everything a connected assistant can reach, so reading before you confirm protects your data.

    Practice: Practice: Connected assistants and prompt injection · 3 questions

Part 6Module 6Approx. 9 min

  1. Checking outputs by type9 min

    Gives a four-part check for any generative AI output, and the weak spots of replies, figures, sources, summaries, translations and code.

    What this module covers

    Starts with a calm, well-written customer reply drafted by an assistant that promised a refund the client had never approved. Explains why confabulation and automation bias make checking essential, and the role of human reviewers in the ISO/IEC 42001 guidance. Sets out Sterling's four-part check: is it true, where did it come from, does it fit, and is it yours to use, with NIST's description of high-integrity information. Covers the weak spot of each output type, false or out-of-date sources, translations, and matching the depth of the check to the stakes, with an insurer in Chicago. Shows how G-H-J's customer service team now checks every draft.

    What you will be able to do

    • Apply the four-part check to any AI output
    • Check the weak spot of each output type, from promises to code
    • Match the depth of checking to the stakes of the work

    Why it matters

    The output you send carries your name and your client's trust, and a minute of checking stops fluent errors reaching them.

    Practice: Practice: Checking outputs by type · 3 questions

Part 7Module 7Approx. 9 min

  1. Intellectual property9 min

    Covers intellectual property and other people's material when using generative AI, both on the way in and on the way out.

    What this module covers

    Opens with a sales manager at G-H-J in Singapore who had an assistant reword pages of a paid consulting report for a pitch, and the complaint that followed. Explains NIST's intellectual property risk and that the law is still being debated, then two questions about rights on the way in and on the way out. Gives Sterling's recommended rules for paid reports, client documents, generated images and code, which are not legal advice. Covers others' confidential material under ISO/IEC 27001 controls 5.32 and 6.6, people's likeness and voice with an architecture firm in Dubai, ownership of outputs, and licences for open-source tools and data.

    What you will be able to do

    • Ask whether you may use material and whether the output copies someone's work
    • Apply the rules for paid reports, client documents, images and code
    • Add your own work and keep your drafts before claiming any rights

    Why it matters

    Rewording someone else's work with a tool does not make it yours, and respecting rights avoids complaints that are hard to answer.

    Practice: Practice: Intellectual property · 3 questions

Part 8Module 8Approx. 9 min

  1. Bias and fairness in outputs9 min

    Explains how bias and unfairness appear in generative AI outputs, where they matter at work, and the rules that keep people deciding.

    What this module covers

    Starts with a recruiter in G-H-J's HR team whose AI shortlist of ten candidates all shared a very similar profile. Explains NIST's description of harmful bias, including under-representation in image models and weaker performance in some languages, and G-H-J's rule on summaries in less common languages. Shows where bias matters most, in decisions and content about people, and the ISO/IEC 42001 guidance on fairness and human oversight. Sets out three rules Sterling recommends, including that AI never ranks, scores or rejects people on its own, habits for content about people, and homogenisation, with a marketing team in New York. Shows how G-H-J's HR team now uses AI.

    What you will be able to do

    • Recognise bias in outputs, including weaker performance in some languages
    • Keep every decision about a person with a named person
    • Ask for variety and check whole sets of content about people

    Why it matters

    A tool with no opinions can still repeat unfair patterns, so keeping people deciding about people protects fairness and trust.

    Practice: Practice: Bias and fairness in outputs · 3 questions

Part 9Module 9Approx. 9 min

  1. Disclosure and records9 min

    Covers when and how to disclose generative AI use, what Article 50 of the EU AI Act says, client expectations and the records to keep.

    What this module covers

    Opens with a bank client asking whether AI wrote a G-H-J monthly report and who checked it, a question nobody could answer. Explains Article 50 of the EU AI Act, applying from 2 August 2026: disclosing deep fakes, AI-made text published to inform the public unless under human editorial responsibility, and telling people they are interacting with an AI system. Gives Sterling's recommended disclosure rule by type of work, and three habits: say it plainly, say it early, and say who checked it. Covers customer expectations under ISO/IEC 42001 control A.10.4, a short record for each client deliverable, and provenance as NIST describes it.

    What you will be able to do

    • Decide when AI use must be disclosed under Article 50 and client terms
    • Disclose AI use plainly, early, and with who checked the work
    • Keep a short record of what AI did on each client deliverable

    Why it matters

    Discovering AI use later damages trust, while planned disclosure and a short record let you answer a client's question in minutes.

    Practice: Practice: Disclosure and records · 3 questions

Part 10Module 10Approx. 9 min

  1. When things go wrong9 min

    Shows what to do when generative AI use goes wrong: what counts as an AI incident, the first five minutes, and learning from it.

    What this module covers

    Starts with a finance analyst at G-H-J's Mumbai centre who uploaded the payroll file by mistake and reported it within ten minutes. Gives NIST's definition of an AI incident, and ISO/IEC 27001 control 6.8 on reporting observed or suspected events. Sets out three steps for the first five minutes, recommended by Sterling: stop, note the facts and report. Covers raising concerns about an AI system under ISO/IEC 42001 control A.3.3, with protection from reprisals, and what the organisation does next: assess, contain, decide who must be told, and learn. Shows the three gaps G-H-J's review found, and why near misses matter, with a hospital in Sydney.

    What you will be able to do

    • Stop, note the facts and report an AI mistake within minutes
    • Report suspected events through your organisation's channel
    • Raise concerns about an AI system and share near misses

    Why it matters

    Deleting the chat and saying nothing lets harm grow unseen, while fast reporting lets the organisation act in time.

    Practice: Practice: When things go wrong · 3 questions

Part 11Module 11Approx. 10 min

  1. Safe use in practice10 min

    Brings the training course together in an eight-line rules card for safe generative AI use, tested on six scenarios from different teams.

    What this module covers

    Closes the training course by returning to the three moments from Module 1 and setting out Sterling's eight-line rules card, from approved tools to reporting. Shows how to explain the rules to a new colleague in two minutes, and three extra jobs for team leaders, with a note from clause 5.1 of ISO/IEC 42001 on modelling responsible use. Tests the rules on six scenarios: a customer reply, a password in an error log, ranking applicants, an AI video of a chief executive, a suspicious connected assistant action and an unapproved app. Gathers five takeaways, shows where G-H-J ended up six months on, and introduces the final exam.

    What you will be able to do

    • Apply the eight rules to everyday AI use at work
    • Explain safe AI use to a new colleague in two minutes
    • Identify which rule applies in a workplace scenario

    Why it matters

    Rules on paper only work when people follow them, and one card everyone knows lets staff keep using AI every day, safely.

    Practice: Practice: Safe use in practice · 3 questions

About this course

Why generative AI needs working rules

Chat assistants, writing help in office software, coding assistants and meeting transcription are now part of everyday work. Each one raises the same three moments: what goes in, what comes out, and what happens next. NIST's Generative AI Profile, NIST AI 600-1 of July 2024, names twelve risks that generative AI creates or makes worse, from confabulation, where confident answers are false, to data privacy, intellectual property, information security and over-reliance. In the European Union, Article 4 of the AI Act, as replaced in July 2026, asks providers and deployers to take measures to support the AI literacy of the people who use AI on their behalf. Article 50, applying from 2 August 2026, adds disclosure duties for deep fakes and for AI-made text published to inform the public.

What makes this course different

The course is a working rulebook rather than an introduction. It follows one built company, GHJ Business Services, with a head office in Singapore and delivery centres in Manila and Mumbai serving clients in the US, the UK, Australia and the EU. Learners see the decisions behind each rule: a never-paste list, a three-colour way to decide for other data, approved tools and business accounts, limits for assistants connected to e-mail and files, a four-part check for outputs, rules for other people's material, people-first rules for decisions about people, disclosure and short records for client work, and the first five minutes after a mistake. Each rule is tied to NIST AI 600-1, ISO/IEC 42001:2023, ISO/IEC 27001:2022 controls or the EU AI Act text, and Sterling's own recommendations are labelled as such.

Who should take it

The course suits any employee who uses generative AI at work, in client service, finance, HR, sales, marketing, analysis or software development, and the team leaders and managers who set rules for them. It is especially useful in service businesses that handle client data under contract, such as business process and shared services centres, consultancies and agencies. No technical background is needed. Learners who want the basics first, such as how generative AI produces answers and how to write good prompts, can start with the Sterling Academy course on AI literacy and responsible use of generative AI; this course does not repeat its case studies.

What a learner can do afterwards

After the course, a learner can name the main generative AI risks in NIST's terms and say which ones meet their own work. They can decide in seconds whether information may go into a tool, cut data down to what a task needs, and use only approved tools and accounts. They can recognise prompt injection and the signs that a connected assistant is being misused, and refuse to confirm an action that looks wrong. They can check an output for truth, sources, fit and rights, keep decisions about people with people, disclose AI use where it is required, keep a short record of AI-assisted client work, and report a mistake or concern quickly. Practice questions follow each of the eleven modules, and a scenario-based final exam leads to the Certificate of Completion.

What each module covers

Module 1, Why safe use needs rules: explains why safe use of generative AI at work needs clear rules, the public sources the course uses, and the three moments that matter. Module 2, The risks NIST names: walks through the twelve generative AI risks NIST names in its Generative AI Profile, and the few that office work meets every day. Module 3, Data you must never paste: sets out the data you must never paste into a generative AI tool, a three-colour way to decide the rest, and how to cut the risk. Module 4, Approved tools, accounts and settings: explains why approved generative AI tools, business accounts and the right settings matter, and what to do when a new AI feature appears. Module 5, Connected assistants and prompt injection: covers AI assistants connected to e-mail and files, direct and indirect prompt injection, and the simple limits that keep them safe. Module 6, Checking outputs by type: gives a four-part check for any generative AI output, and the weak spots of replies, figures, sources, summaries, translations and code. Module 7, Intellectual property: covers intellectual property and other people's material when using generative AI, both on the way in and on the way out. Module 8, Bias and fairness in outputs: explains how bias and unfairness appear in generative AI outputs, where they matter at work, and the rules that keep people deciding. Module 9, Disclosure and records: covers when and how to disclose generative AI use, what Article 50 of the EU AI Act says, client expectations and the records to keep. Module 10, When things go wrong: shows what to do when generative AI use goes wrong: what counts as an AI incident, the first five minutes, and learning from it. Module 11, Safe use in practice: brings the training course together in an eight-line rules card for safe generative AI use, tested on six scenarios from different teams.

Frequently asked questions

Your questions,
answered.

About Safe Use of Generative AI at Work Training, its certificate and enrolment.

Ask Sterling
Is Sterling Academy a good choice for Safe Use of Generative AI at Work training?

Yes. Sterling Academy's Safe Use of Generative AI at Work training is built by the Sterling consulting group, which has taken organisations to certification across 8,100+ client engagements in 60+ countries over 20+ years. It comes with practice questions that show the right answer. Each course ends with a final exam and a verifiable certificate that anyone can check online, plus a LinkedIn badge, and client administrators can show an auditor exactly who has been trained. You will be able to apply Safe Use of Generative AI at Work at work and show your competence with confidence. Sterling Academy courses carry the same content as Sterling's classroom training, whose certificates are trusted and accepted by all major certification bodies, as evidence of training at client audits for over 20 years.

Who should take safe use of generative AI at work training?

It is for any employee who uses a chat assistant, a writing or coding assistant, or AI features in office software, and for the team leaders and managers who set rules for them. It suits client service, finance, HR, marketing, analysts and developers. No technical background is needed: each rule is explained with real working situations.

How is this course different from AI Literacy and Responsible Use of Generative AI?

The AI literacy course explains what generative AI is, how to prompt well and why outputs can be wrong. This course is the next step: the working rules for safe use, based on the NIST Generative AI Profile, covering data classification, approved tools and settings, prompt injection, checking by output type, intellectual property, disclosure, records and reporting.

How are learners assessed in the safe use of generative AI at work course?

Short practice questions after each module help you check your understanding; they do not count toward the result. The course ends with a scenario-based multiple-choice final exam. Learners who score 60% or more receive a Sterling Academy Certificate of Completion, and one retake is included. Each certificate carries a unique number and a QR code for online checking.

Does the EU AI Act require AI training for staff?

Article 4 of the EU AI Act, as replaced by Regulation (EU) 2026/1744, requires providers and deployers of AI systems to take measures to support the AI literacy of their staff and others using AI systems on their behalf. It does not require any specific level for each person. The course explains the article in plain words.

When will Safe Use of Generative AI at Work Training be available?

The course is in preparation. Register on this page with “Notify me when it opens” and we will send you one e-mail when it opens for enrolment, with the price per learner and the details for team enrolment. The course page will then show the full module list and the video runtime.

Your certificate

Prove what you learned.
Share it with confidence.

You will be able to apply Safe Use of Generative AI at Work at work and show your competence with confidence. Your certificate can be verified online in seconds, so certification body auditors, employers and clients can confirm your training instantly.

Complete the course and its final exam and you receive a Sterling Academy Certificate of Completion, verifiable online by QR code, with a LinkedIn badge to share.

Pass mark
60%
Prerequisites
None

Sterling Academy courses carry the same content as Sterling's classroom training, whose certificates are trusted and accepted by all major certification bodies, as evidence of training at client audits for over 20 years.

Ask about course requirements

Your next step

Get consulting from Sterling

Expert and experienced consultants, 8,100+ client engagements in 60+ countries, from gap assessment to a system ready for its certification audit.

Talk to a Sterling consultant

Not sure which course?

Plan your next step
in learning.

Discuss your learning needs