Loading…

Risks: hallucinations, copyright, privacy and brand safety

5 minutes, 8 steps, workout 6 of 6 in this module

An AI assistant writes a product description in seconds, and nobody notices that it promises a feature the product doesn't have or reuses a sentence from a competitor's site. The first person to catch it might be a customer, a lawyer or a journalist. Marketers who publish AI output without a review step end up spending far more time apologizing than they saved, so knowing where these tools go wrong is part of using them at work.

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

  1. 1The four risks: hallucination, copyright, privacy, brand safety
  2. 2Case: 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?

  1. Nothing, if it sounds plausible customers won't check
  2. The assistant invented a fact. The claim has to be backed by evidence you actually have, or deleted
  3. Only the number is wrong; change 40% to 20% and it's fine
  4. 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