Section 05
Risk Examples & Critical Thinking
None of this is a reason to avoid AI. It's the stuff that separates entrepreneurs who use it well from entrepreneurs who get caught out — usually in front of an audience.
Five ways it goes wrong
Each one ends with a question to sit with — not a rule to memorise.
Related terms: , , , .
It is increasingly important to stay informed about new regulatory frameworks, such as the EU AI Act or region-specific privacy laws, which shape how AI should be deployed responsibly. By being mindful of how we interact with AI, we can prevent unintended consequences like reinforcing biases, infringing on privacy, or eroding trust in technology.
Europe sorts AI uses by how much harm they could do. Most of what you'll build sits in the green or yellow band — but knowing the map makes you sound like someone who's done their homework.
Minimal risk
Green band
No special obligations. Use freely, your own judgement still applies.
- — Using AI to draft your Instagram captions
- — Spam filters and recommendation features
- — Brainstorming brand names for your startup
Limited risk — transparency required
Yellow band
People must be told they're dealing with AI or AI-generated content.
- — A chatbot on your website needs to say it's a bot
- — AI-generated images or video in your ads need to be labelled
- — Deepfake-style content must be disclosed
High risk — human accountability
Orange band
Strict duties: risk management, data quality, documentation and meaningful human oversight. A person stays accountable for the outcome.
- — AI screening job applicants for your startup
- — AI used in education to score exams or grant access
- — Credit scoring and access to essential services
Unacceptable risk — prohibited
Red band
Banned in the EU. Not a grey area.
- — Social scoring of people by public authorities
- — Manipulative systems that exploit children or vulnerable people
- — Untargeted scraping of faces to build recognition databases
What happens if you break it
The AI Act sets maximum fines of up to €35 million or 7% of worldwide annual turnover (whichever is higher) for prohibited practices; up to €15 million or 3% for most other breaches; and up to €7.5 million or 1% for supplying incorrect or misleading information to authorities. Smaller companies face the lower of the two figures rather than the higher.
Not legal advice — check the official AI Act text (linked in Resources) for anything that affects a real business.