Keeping it safe · for marketers
Using AI safely in marketing
Marketing teams handle customer lists, campaign results, unreleased product plans, and partner agreements. Much of that should never go into a tool that has not been approved for it. The risk is not only a leak. It is also a regulatory problem with customer data, or a breach of an agreement with a partner.
Keeping it safe also covers the output. Content that copies a competitor too closely, invents a testimonial, or makes a claim the company cannot support creates legal and reputational risk.
Where it comes up
Analyzing a customer list
Before any segmentation work, remove names, emails, and other identifying fields, or use a tool your company has approved for customer data.
Drafting launch copy before a launch
Unreleased product details are confidential. Use only the approved tools for them, and check what your company’s AI policy allows.
Generating testimonials or reviews
This is a task to refuse. AI can help you ask real customers for quotes or edit a real quote with permission. It must not invent one.
What it looks like at each band
E1 · Foundational
Knows that private data should not go into AI tools, but does not always recognize it in a spreadsheet or a pasted email thread. Follows the policy when reminded of it.
E2 · Proficient
Removes identifying customer data before use as a habit. Uses approved tools for confidential material. Recognizes content that should not be AI-generated, such as testimonials or legal claims.
E3 · Distinguished
Judges risk across a whole campaign, including data, confidentiality, claims, and imagery rights. Raises concerns early rather than at sign-off. Knows when a task should not use AI at all.
E4 · Exceptional
Shapes the team’s practice: which tools are approved for which data, what review AI content needs, and how disclosure works where it is required. Handles a mistake openly and fixes the process that allowed it.
Common mistakes
Pasting a raw export
CRM and email platform exports carry names, emails, and purchase history. Strip the identifying columns first.
Invented social proof
A testimonial, review, or customer count that AI produced is fabricated, however realistic it reads. It can breach advertising rules and the trust of your audience.
Close imitation of a competitor
Asking AI to write "like" a named competitor’s campaign can produce copy too close to theirs. Brief on your own positioning instead.
A worked example
Finding themes in 2,000 rows of post-purchase survey responses.
Upload the full survey export, with names and emails, to a personal AI account and ask for the main themes.
Customer personal data has gone into a tool the company has not approved, which may breach the privacy policy customers agreed to.
Delete the name, email, and order ID columns. Confirm that the remaining free-text answers do not contain personal details. Use the company’s approved AI tool. Ask for themes with the number of responses behind each, then spot-check the themes against the raw answers.
The analysis still gets done. Customer data is protected, the tool is approved, and the themes are checked against the source.
The full study guide for this skill, for every role, is Keeping it safe. The other five skills for marketers are on the AI skills for marketers page. A one-page summary is in the AI data safety checklist.
See where you stand
The free check grades keeping it safe and the other five skills in about fifteen minutes. Nobody sees the grade but you.
Then build the skills and prove it. The three steps for marketers list the courses and templates that fit the work.