Loading…
Risks: hallucinations, copyright, privacy and brand safety
5 minutes · 8 steps · workout 6 of 6 in this module
What you'll be able to do
By the end of this workout, you'll be able to name the four main risks of AI in marketing, put a cost on a false claim, and set up a review step that catches problems before they're published.
A first look: The four risks: hallucination, copyright, privacy, brand safety
Hallucination: the assistant states things that aren't true, in a confident tone, with made-up numbers, studies, quotes or product features. Copyright: images or text that copy a living artist's style, a competitor's copy, a trademark or a real person's face; and the terms of the tool you use, which decide what you're allowed to publish commercially. Privacy: customer names, emails, addresses and order details pasted into a consumer AI tool may be stored or used for training; data-protection rules apply to what you paste, not just to what you publish. Brand safety: content that's off-tone, biased, or fake, like generated reviews, or that platforms require you to label as AI-made. One principle covers all four. People weigh a broken promise far more heavily than a kept one. That's loss aversion: one false claim costs more trust than ten good posts earn.
Example: Corner Coffee's rules card, pinned above every desk: no number without a source, no style-of prompts, no customer data in prompts, no post without a person's name on the approval.
What this workout covers
- The four risks: hallucination, copyright, privacy, brand safety
- Case: Northpack's support inbox and the chatbot
A question from this workout
An AI assistant drafted this line for Green Garden Skincare: “Clinically proven to reduce wrinkles by 40% in two weeks.” The brand has never run a clinical study. Before we go into the risks, what's the problem here?
- Nothing, if it sounds plausible customers won't check
- The assistant invented a fact. The claim has to be backed by evidence you actually have, or deleted
- Only the number is wrong; change 40% to 20% and it's fine
- It's the assistant's responsibility, since it wrote the sentence
Pick your answer first, then open the reasoning below.
Show the answer and the reasoning
Answer: B. The assistant invented a fact. The claim has to be backed by evidence you actually have, or deleted
This is a hallucination: the assistant produced a confident, specific, false claim because that's what skincare ads tend to sound like. A claim you can't back up is a legal problem and a trust problem, whatever the number. The most common mistake is option c: softening the number. A smaller invented number is still invented. The person who publishes is responsible, not the tool.
The other steps work the same way: you decide first, then see why each option is right or wrong.
How you practice here
- 1 short concept card
- 2 multiple-choice questions
- 1 campaign case with real-looking numbers
- 1 put-in-order exercise
- 1 matching exercise
- 1 guided calculation
- 1 recall question from an earlier workout
Part of the module: AI in Marketing
Make AI your teammate for content, ad creative, analysis and research, and understand how AI search is changing marketing.
More workouts in this module
- Writing good prompts: role, context, examples, format
- Creating content with AI: speed, brand voice and control
- AI for ad images and video creative
- AI for data analysis: from spreadsheet to insight
- AI search: how ChatGPT, Gemini and AI Overviews are changing traffic
Programs that include this workout