How do you use AI in communications without inventing facts or leaking news?
AI tools can research a story, summarise coverage and draft a release. They can also invent a quote, cite a book that does not exist, or receive information that should never leave the company.
In May 2025 a syndicated newspaper insert ran a summer reading list in which 10 of 15 books did not exist; the writer had used AI and not checked. In a case reported in May 2024, fraudsters used a voice clone of WPP’s chief executive in a Teams meeting to push an agency leader to hand over money and personal details. The attempt failed.
This course gives communications, PR and investor relations teams one method for each of those risks. Brief the tool before it drafts, check every fact against its source, label where the AI Act requires it, keep inside information out of public tools, and record who approved what. By the end of the day the team has a prompt library and a disclosure checklist for its next release.
After this course
- Can you brief an AI tool with approved facts, audience, angle and limits before it drafts a release or pitch?
- Can you check AI research and media monitoring against the original sources before anyone uses it?
- Can you decide when AI-generated text needs a label under AI Act Article 50(4), and keep the review record?
- Can you keep inside information out of public AI tools, applying MAR Articles 7 and 10?
- Can you verify requests that seem to come from executives through a known channel before acting?
- Can you write factual, sourced pages that AI crawlers can read without running JavaScript?
Skills you practise
- Brief-first drafting
- Source checking
- Media monitoring
- AI content labelling
- Inside information handling
- Deepfake verification
- AI search visibility
- Crisis drafting
Course structure
The course is 1 day of 6 teaching hours, in 2 blocks of 3 hours with a lunch break. Online, it runs as 2 live sessions of 3 hours on the same day, not recorded.
The first block covers daily work. Participants run research and media monitoring prompts with search switched on, then check each claim and link against the original. They write a brief before any draft: approved facts, audience, angle, structure, length and tone, with gaps marked for confirmation.
The block ends with labelling. From 2 August 2026, Article 50(4) of the EU AI Act requires deployers to disclose AI-generated text published to inform the public on matters of public interest. The duty does not apply where the text had human review and someone holds editorial responsibility. Participants set the team’s rule and the record that shows the review.
The second block covers higher-risk work. Under Article 7 of the EU Market Abuse Regulation (MAR), inside information is information of a precise nature, not made public, about an issuer or its financial instruments, that would be likely to have a significant effect on prices if made public. Article 10 makes it unlawful to disclose inside information to another person outside the normal exercise of an employment, a profession or duties.
We found no statement from ESMA, the EU securities markets authority, or from the Dutch AFM on staff typing inside information into AI tools (checked 15 October 2026). ESMA’s May 2024 statement and the AFM’s April 2026 report on AI in capital markets cover other AI uses. The course treats it as a material, untested risk and keeps inside information out of public tools.
Participants then look at how AI answer engines read the organisation’s own pages. A December 2024 Vercel study found that the AI crawlers of OpenAI, Anthropic and Perplexity did not run JavaScript, so key facts belong in plain, structured HTML with sources.
The day ends with a crisis drill: a fake video of the chief executive, a holding statement in 3 versions, a verification rule for executive requests, and the approval chain on record.
About two thirds of the time is hands-on, in the tool your organisation has approved.
1Research and monitoring you can check60 min
- Desk research with search switched on
- Summaries of media and social coverage
- Made-up quotes, books and sources
- Checking each claim against the original
A model can produce a quote, a figure or a source that does not exist, in the same confident tone as a correct one. Every fact that leaves the team needs a source someone opened.
Caution. In May 2025 a syndicated newspaper insert listed 15 summer books, 10 of which did not exist, after the writer used AI and did not check. Open every source before you quote it.
You leave with: A monitoring prompt the team can rerun each week, with the checks that go with it.
2Releases and pitches, brief first75 min
- The brief: approved facts, audience, angle, limits
- Press releases from a fact sheet
- A pitch written for one journalist
- House style and voice examples
A language model writes the most likely next words. Without your facts and your angle, the most likely press release is a generic one, so the brief comes before the draft.
Caution. In October 2025 Deloitte Australia refunded part of its fee for a government report that contained an invented quote from a court judgment. Every quote in a release needs written approval from the person quoted.
You leave with: A brief template and tested release and pitch prompts for the prompt library.
3Labelling AI content under the AI Act45 min
- Article 50(4): AI text on matters of public interest
- The human review and editorial responsibility exemption
- Labelling deepfake images, audio and video
- Dates and fines from 2 August 2026
The editorial exemption in Article 50(4) for AI text on matters of public interest needs human review or editorial control, and a natural or legal person who holds editorial responsibility. Record each review with the editor's name and the date.
Caution. From 2 August 2026, breaches of Article 50 can be fined up to €15 million or 3% of turnover. Keep the review record for every AI-assisted text you publish on matters of public interest.
You leave with: A labelling rule for the team: when to label, when the review exemption applies, and which record to keep.
4Inside information and public AI tools60 min
- Inside information under MAR Article 7
- Unlawful disclosure under MAR Article 10
- Inside information in public AI tools: no ESMA or AFM statement yet
- Drafting results releases with placeholders
Text pasted into a public AI service leaves your organisation the moment you send it. Keep the figures that make a release inside information out of the chat.
Caution. We found no ESMA or AFM statement on staff entering inside information into AI tools. Treat it as a material, untested risk under MAR Article 10, and keep inside information out of public tools.
You leave with: The inside information section of the disclosure checklist, and a placeholder prompt for results and deal releases.
5How AI answer engines describe your organisation45 min
- How answer engines fetch and read pages
- AI crawlers that do not run JavaScript
- Factual, structured pages with sources
- Checking what answer engines say about you
A model knows only what was in its training data unless it searches the web when it answers. What an answer engine says about your organisation depends on the pages it could fetch and read.
Caution. In a December 2024 Vercel study, the AI crawlers of OpenAI, Anthropic and Perplexity did not run JavaScript, so put key facts in the page HTML where those crawlers can read them.
You leave with: A facts page draft and a monthly prompt that checks what answer engines say about the organisation.
6Fake executives, crisis drafts and approvals75 min
- Voice clones and fake accounts of executives
- Verification questions and call-back rules
- Holding statements for fake video or audio
- Who approves what, and the record of it
In a crisis the tool drafts and a named person approves every word that goes out. Nothing is published automatically, however urgent it looks.
Caution. In a case reported in May 2024, fraudsters used a voice clone of WPP's chief executive and his photo on WhatsApp to target an agency leader. Confirm any request that seems to come from an executive through a known number or a pre-agreed question.
You leave with: A verification rule for executive requests, a holding statement prompt set, and the completed disclosure checklist.
Research and monitoring you can check
Ask the tool to search, link every item and mark what it could not confirm.
What are journalists saying about our sector this month?
Search news published in the last 30 days about [sector] in [country]. List the 8 most relevant articles as a table: date, outlet, journalist, headline, one-line summary, link. Include only articles you opened through search. Mark any field you could not confirm as 'unconfirmed'.
Releases and pitches, brief first
Name the facts, audience, angle, structure, length and tone before asking for a draft.
Write a press release about our new product launch.
Write a press release from the fact sheet below. Audience: trade journalists in logistics. Angle: the product reduces picking errors in warehouses; use only the figure in the fact sheet. Structure: headline under 12 words, 3 paragraphs, 1 quote from our CEO marked [QUOTE TO APPROVE], company boilerplate. Length: under 400 words. Tone: factual. Mark any missing fact as [TO CONFIRM]. [fact sheet]
Labelling AI content under the AI Act
Ask for a numbered list of every change between the AI draft and the edit.
Make this AI draft sound more human so we don't need to label it.
Compare the AI draft and the edited version below. List every factual change, every added or removed claim, and every change in meaning as a numbered table with 3 columns: AI draft, edited version, type of change. I will attach the table to the approval record with the editor's name and date. [AI draft] [edited version]
Inside information and public AI tools
Use placeholders for every figure and fact that is not yet public.
Here are our Q3 numbers before Thursday's release: revenue down 9%, guidance cut. Write the press release.
Draft the structure of a quarterly results release for a listed company, using placeholders only: [REVENUE], [CHANGE %], [GUIDANCE]. Sections: headline, 3 key figures, CEO quote marked [QUOTE TO APPROVE], outlook, financial calendar, contacts. Length: under 450 words. Tone: factual. I will add the figures outside this tool after approval.
How AI answer engines describe your organisation
Specify the sections, the style and the source for every fact.
Write an About page that AI search will pick up.
Draft a company facts page from the approved facts below. Sections: what we do (2 sentences), key figures with the date and source of each, leadership with roles, locations, press contact. Style: plain statements, one fact per sentence. End with a list of sources as links. [approved facts]
Fake executives, crisis drafts and approvals
Set the length and format for each channel in one prompt.
Write something saying the CEO video is fake.
Draft a holding statement about a fake video of our CEO circulating on social media. Facts: the video is not authentic; we have reported it to the platform; the CEO made no such announcement. Write 3 versions: a post under 280 characters, a 60-word statement for journalists, and a 2-line message for staff. Tone: calm and factual. Mark every fact that legal must confirm as [CONFIRM].
The team keeps a communications prompt library: tested prompts for monitoring, research, releases, pitches, facts pages and holding statements, each with the checks that go with it.
It also keeps a disclosure checklist with 3 questions for every release. Does the text contain inside information? Does it need an AI label, or does the review exemption apply? Who approved it, and when?
In the week after, the checklist goes into the release approval process and the prompt library goes into the team’s shared drive. Each participant receives an attendance record for your AI literacy file.
- Teaching time
- 6 hours on 1 day, in 2 blocks of 3 hours with a lunch break.
- Before the day
- Participants use an AI assistant at least weekly, or have taken AI Basics at Work.
- After the day
- A communications prompt library for the team and a disclosure checklist covering inside information, AI labelling and approvals.
Completion
You complete the course by working through each exercise on your own tasks. There is no exam and no grade. Each participant receives an attendance record for your AI literacy file (not a certificate).
Who is this course for?
Communications and PR teams, investor relations officers, spokespeople, and the people who approve what they publish. Participants should already use an AI assistant at least weekly.
The inside information module is written for listed companies. For organisations that are not listed, the same module covers other confidential information, such as unannounced deals and results.
Choose AI for Video Production if your team makes video and needs the rules on consent and likeness. Choose Shadow AI and Safe AI Use if the wider organisation still needs rules on which tool may take which data.
This course is not part of a path. Book it on its own.
- Level
- Intermediate
- Before you start
- Participants use an AI assistant at least weekly, or have taken AI Basics at Work.
- Group
- 1 to 15 people, online, live or at your office
This course is right for you if you
- write releases, pitches or investor updates
- use an AI assistant at least weekly
- work for a listed company or handle news that is not yet public
Unsure whether it fits? Ask us. We reply within 1 working day.
Practical information
- 1 day, 6 teaching hours, with a lunch break
- Online: live in 2 sessions of 3 hours on one day, not recorded; times agreed in the intake call
- At your office: 09:30 to 16:30; you provide a room, a screen and internet access
- In English, in the AI tool your organisation has approved
- Taught live by a practitioner with more than 15 years of commercial and technical leadership in software and data companies across Europe and Asia-Pacific, including maritime, energy and fintech.
- Attendance record for your AI literacy file
- Online, live: 1 day in 2 live sessions of 3 hours, not recorded. Book it as a private day for your team of 1 to 15 people, on your own Teams, Zoom or Meet.
- At your office: 1 day (6 teaching hours) for 1 to 15 people, travel within the Netherlands included.
Dates and booking
Choose the format, an open weekday in the calendar and the number of participants, then reserve on this page.
The calendar and booking form load with JavaScript. Without it, ask us for a date on WhatsApp, by phone on +31 6 20 51 73 96 or at advisory@ai-business.school.
Investment
| Item | Online, live | At your office |
|---|---|---|
| Price | €399 per person | €2,900 for the day, for your group |
| Including 21% Dutch VAT | €482.79 per person | €3,509 |
| The training day | Included | Included |
| A 30-minute intake call on your tools and data rules | Included | Included |
| Course materials and templates | Included | Included |
| Attendance record for your AI literacy file | Included | Included |
| Travel within the Netherlands | Not needed | Included |
| Group | 1 to 15 people | 1 to 15 people |
Prices exclude VAT. Online, live, you pay per participant. At your office, one price covers your whole group, from 1 to 15 people; for a larger group, book a second day. You reserve on this page; the payment details appear when you do.
Questions
Can our communications team use ChatGPT or Copilot to write press releases?
Yes, in a tool your organisation has approved and with a brief that holds the approved facts, the audience and the limits. Every quote and figure is checked against its source and approved before release. The course builds both steps into the team's prompt library.
Do AI-written press releases need a label under the EU AI Act?
From 2 August 2026, Article 50(4) requires disclosure of AI-generated text published to inform the public on matters of public interest. The duty does not apply where the text had human review and someone holds editorial responsibility. The course sets the team's rule and the record that shows the review.
Can investor relations staff use AI tools with inside information?
Keep inside information out of public AI tools. MAR Article 10 makes disclosure of inside information outside the normal exercise of duties unlawful, and we found no ESMA or AFM statement on how that applies to AI tools. Treat it as a material, untested risk.
How do we get AI answer engines to describe our company correctly?
Publish factual, sourced pages with the key facts in the HTML, because a 2024 study found that major AI crawlers did not run JavaScript. Then check each month what the answer engines say and correct your own pages where they are unclear.
Is this course suitable for spokespeople?
Yes. Spokespeople join the module on fake executives and crisis drafts, which covers a verification routine for requests that seem to come from executives and holding statements for fake video or audio.
Can we run AI training for a communications team online?
Yes. The online version runs live on your own Teams, Zoom or Meet in 2 sessions of 3 hours, in English.
Last reviewed 20 October 2026.
Related courses
AI for Video Production
Make video with AI tools, with written consent for every face and voice, deepfake labels and every claim checked.
€399 per person online · €2,900 a day at your office, up to 12 people
Shadow AI and Safe AI Use
Use AI at work without leaking data or falling for deepfake scams, and leave with the input for your AI policy.
€399 per person online · €2,900 a day at your office, up to 15 people
AI Literacy and the AI Act
Explain what the EU AI Act asks of your organisation and keep the records for your AI literacy file.
€399 per person online · €2,900 a day at your office, up to 15 people