LANDMARK · 5 MIN · JUDGE

Escalate responsibly

After this landmark, you can recognise the situations where an AI-assisted decision needs to stop and involve legal, privacy, security, or a domain expert — before you act, not after.

EU context
SHARED FOUNDATION

Literacy isn’t knowing every answer; it’s knowing the edge of what you should decide by yourself. Some AI-assisted decisions carry consequences that a confident model can’t bound: a possible legal exposure, personal or regulated data, a security question, a medical or financial call, anything affecting someone’s rights or safety. In those cases the responsible move is to escalate — to pause and bring in the person or function who owns that risk. A model will happily produce a fluent answer to a question that actually requires a lawyer, a doctor, or a security review, and it won’t flag that it’s out of its depth. Escalation isn’t failure or slowness; it’s the checkpoint that keeps a fast tool from committing an organisation, or a person, to something no one qualified reviewed.

This is the canonical concept. It stays the same across learner lenses so personalization never changes the underlying facts.

EVERYDAY LENS

What this looks like for you

Treat AI as a starting point, not the final word, on anything with real stakes — legal, medical, financial, or safety. When a decision could genuinely harm you or someone else if it’s wrong, get a qualified human before acting on the model’s answer.

MAKE A DECISION

A team wants to use an AI tool to screen job applicants and auto-reject low scorers. Before turning it on, what’s the responsible step?

CARRY THISList three decisions in your work where you’d stop and escalate, and name the person or function you’d escalate each to.
PRACTICE THIS
CONCEPTS IN THE INDEX