Purpose

This memo provides practical ethical guidance for organisations using customer data in AI systems. It is not legal advice — it is a thinking framework to ensure your AI practices align with customer trust and societal expectations.

Core Principles

1. Consent Must Be Meaningful

Burying AI data usage in page 47 of your terms of service is legally defensible but ethically insufficient. Customers should understand, in plain language, how their data informs AI systems.

In Practice

Consider implementing tiered consent: basic service vs. AI-enhanced features. Let customers choose their level of participation.

2. The Reasonable Expectation Test

“Would our customers be surprised to learn we use their data this way?” If the answer is yes, you have an ethical gap regardless of what your privacy policy says.

Customer support transcripts used to improve support AI? Reasonable. The same transcripts used to predict churn and trigger retention offers? Less clear.

3. Data Minimisation Is an Ethical Imperative

Collect and process only what you need. The temptation with AI is to hoover up everything because “the model might find patterns.” This approach creates risk without proportionate benefit and erodes the trust relationship.

4. Anonymisation Is Harder Than You Think

True anonymisation in the age of AI is extremely difficult. Re-identification attacks are increasingly sophisticated. Treat “anonymised” data with nearly the same care as personal data, especially when combining datasets.

5. The Right to Explanation

When AI systems make decisions that affect customers, those customers deserve to understand why. “The algorithm decided” is not an acceptable explanation. Build explainability into your systems from day one.

Recommended Actions

Conduct a data ethics audit of all customer-facing AI systems within 90 days
Update customer communications to clearly explain AI data usage
Establish a process for customers to opt out of AI processing
Document the ethical reasoning behind each AI use case

Ethics Memo: Using Customer Data in AI Systems