AI Literacy Series: Where Your Prompts Go: How Providers Use Your Information, Privacy Law, Professional Codes and Safe AI Use with Client Work

When you paste a client email, a contract or a set of figures into an AI tool, a readable copy lands on someone else's computers. What happens to it next depends on the account you are signed into, the settings you have chosen, and the legal and professional rules you work under. This course follows that copy step by step. 

Dr. Jason Dean starts with the route a prompt takes after you press Enter and the four things that can happen to the copy: training, human review, safety logging and legal holds. He then compares, in plain language, how ChatGPT, Claude, Gemini and Microsoft Copilot treat personal accounts and business accounts, including retention, human access, feedback buttons, memory and share links. Real settings screens show where the training controls are. 

The course then separates documented incidents, such as search-indexed share links and the Samsung source-code case, from fears that the public record does not support, and explains shadow AI, legal holds and data residency. The legal section uses Canada as the worked example and shows how the same question is handled under the EU and UK GDPR, United States sector rules, Australia's Privacy Act and professional codes such as the IESBA Code and ABA Formal Opinion 512. It closes with practical tools: a data ladder, a green-amber-red sorting rule, a de-identification exercise, sample client-disclosure wording and seven working habits. 

By the end of the course, learners should be able to: 

- trace where a prompt goes after it is submitted and identify which copies their own settings control;

- compare how major AI providers treat personal and business accounts for training, retention, human review and sharing;

- explain why account type, settings, agreements and data location change what a provider may do with information;

- distinguish documented AI privacy incidents from unsupported fears, and recognise shadow AI;

- apply privacy-law and professional-confidentiality principles to AI use, using Canada as a worked example and recognising how other jurisdictions differ;

- sort a work task green, amber or red using sensitivity, account and task; and

- de-identify a prompt while recognising the limits of de-identification.