When should you stop or avoid using AI?
Deciding when not to use AI starts with the law but does not end there. Some uses are banned in the EU, and others are lawful but should be paused because their risks are not under control.
The legal red lines
Legal requirement (EU AI Act, Art. 5). The AI Act lists prohibited practices: uses of AI considered so harmful that they are not allowed at all (AI Act, Art. 5). These bans have applied since 2 February 2025. From 2 December 2026, the Digital Omnibus on AI adds new prohibitions on generating child sexual abuse material and non-consensual intimate deepfakes (Regulation (EU) 2026/1744). The full list and timeline, in plain language, are in what the EU AI Act requires. If a use case falls under a prohibition, the answer is to stop it, not to adjust it.
Signals to pause
Recommendation. Outside the prohibitions, pause a use case, or do not start it, when one of these signals appears:
- The output cannot be explained or controlled. Nobody can say why the system produces a given result, check it before it takes effect or stop it quickly.
- The harm outweighs the benefit. A wrong output could seriously affect someone’s job, money, health or rights, while the gain is mainly speed or convenience. The voluntary OECD AI Principles are a useful reference for this weighing.
- There is no lawful basis for processing the data. Under GDPR, personal data may be processed only on a legal basis (GDPR). If the tool needs personal data that you have no lawful basis to use, the use case cannot go ahead in that form. Unlike the other two signals, this one reflects a legal requirement that already applies.
A pause is a decision, not a punishment. Name who makes it, what has to change before the use case can resume and who checks that it has.
“Do not use AI” is a valid result
Knowing when not to use AI is part of using it responsibly. An AI risk assessment that ends with “not here” or “not yet” has done its job. Record the decision and the reasons, so the same idea is not reopened from scratch six months later without anything having changed.
Next step: go through your AI inventory and flag every use case that shows one of these signals. For each one, decide whether to pause, change or continue it, and write down who decided.
Sources and further reading
- Regulation (EU) 2024/1689 (AI Act) — Article 5
- Regulation (EU) 2026/1744 (Digital Omnibus on AI)
- Regulation (EU) 2016/679 (GDPR)
- OECD AI Principles
This article is for general information and is not legal advice.
AI Horizon Conference
The AI Horizon Conference returns to Lisbon, once again bringing together entrepreneurs, investors and industry leaders to discuss the future of AI.