Academy

AI Governance as a User

What AI is, where it fails, what you owe, and when to escalate. Written for anyone whose work now involves AI, whether or not that was ever in the job description. Supports the AI literacy duty that the EU AI Act places on providers and deployers, proportionate to role.

Sign in to start

  1. What is AI

    0/1

    Where the boundary sits between AI and ordinary software, and why the legal definition matters more than the marketing one.

  2. How machine learning works

    0/1

    Training and inference, and why accuracy on a test set is not accuracy in your hands.

  3. Generative AI and large language models

    Next-token prediction, and why fluency and correctness are unrelated properties.

    • AI agents and automation

      What changes when a model can plan, remember, call tools and act.

      • Where AI fails

        Failure modes as a taxonomy rather than a list of anecdotes.

        • Bias and discrimination

          How bias enters, why fairness metrics are incompatible, and what the law prohibits.

          • Hallucination and unreliability

            Why confident fabrication is intrinsic, and the verification habits that survive deadlines.

            • Data protection and the privacy perimeter

              What leaves your organisation when you paste into a chatbot.

              • Security risks

                Prompt injection, data leakage, and what an attacker does to an AI feature.

                • Synthetic content and the rules around it

                  Disclosure, visible labelling and machine-readable marking as three separate duties.

                  • When AI is used on your data

                    The rights an individual has when a significant decision is automated.

                    • AI at work, and the AI you do not see

                      AI already embedded in the tools you use, and how features arrive by vendor update.

                      • Explainability and transparency

                        What an explanation is worth, and the difference between an explanation and a justification.

                        • High-quality prompting

                          Specification rather than incantation, and why long prompts are not better prompts.

                          • The AI supervisor

                            Human oversight as a statutory concept, and why approving at volume is not oversight.

                            • The escalation matrix

                              What to escalate, to whom, how fast, and when the reporting clock starts.

                              • AI governance: what it means for each of us

                                Governance as individual duties rather than a committee's problem.

                                • Working with an AI policy, or without one

                                  How to read a policy for what it actually requires, and what to do when there is none.

                                  • Governance platforms and shadow AI

                                    What these platforms do and do not do, and how unapproved use gets discovered.

                                    • AI in business and governance workflows

                                      Where AI sits in the work, and the risk of governing an opaque system with another one.