Course 5 of The AI Decision Course
Governance, risk and responsible AI
For leaders who answer for what AI does: what to check before it speaks in public, vendor terms, bias, disclosure, and safety work before release.
26 short lessons. Each one is a five-minute read with a card your team fills in and three AI prompts to test it.
Lessons in this course
- Count the AI you already have
- Match the human check to the harm a mistake can do
- How much AI governance to build before anything scales
- Name the person who answers for each agent
- Whoever uses the AI answers for what it wrote
- A check before AI speaks to the public
- Ban the tools, or give staff a tool of your own?
- Four questions that decide what you tell people about AI
- Fakes made with your name, face or voice
- Five places an AI system picks up bias
- Average accuracy hides the groups your AI fails
- Strip the identifiers before AI touches the data
- Where the line falls on data about people
- Five tests to run on an AI vendor before you sign
- Size the human check, then shrink it
- Set how much explanation each AI use has to give
- Can you own what AI makes for you?
- Five checks on a model's training data before it goes into the business
- Who carries the copyright risk when you publish AI output
- Compliance that holds when the rules move
- Planning an AI launch when the rules change at the border
- Decide the safety work before the release date decides it
- Governing agents all the time, not twice a year
- How far you answer for what people do with your AI
- Regulated products: making your evidence survive the next model version
- Four questions to put to a call for AI rules