Responsible Use of Generative AI

Knowing how these models work and where they fall short (the last seven lessons) is what makes it possible to use them responsibly. This final lesson covers the practical side of that.

Attribution and copyright questions

Generated text, images, and code are produced from patterns learned across enormous training datasets, and the legal questions around ownership, licensing, and attribution of that output are genuinely unsettled and vary by jurisdiction and provider terms. When it matters — publishing commercially, using generated code in a product, using generated images publicly — check the specific provider's terms of use rather than assuming a default answer.

Disclosing AI-generated content

Many contexts (academic work, journalism, some platforms' terms of service) require disclosing when content was AI-generated or AI-assisted. Passing generated content off as entirely your own unedited work, where disclosure is expected, is a trust and integrity problem independent of whether the content itself is any good.

Don't treat output as ground truth

This connects directly back to lesson 7: an LLM's fluent tone is not a signal of accuracy. Before acting on generated content — publishing it, running generated code, using a generated fact in something important — review and verify it the same way you would a first draft from a person you don't know yet.