The EU AI Act (phasing in from 2024-2026) classifies AI systems by risk. For designers, the relevant distinction is between minimal/limited and high risk: healthcare, HR, credit, education, justice, critical infrastructure. In these domains, transparency and human oversight requirements become mandatory — and translate into interface choices.
Obligations impacting UX: inform that the user interacts with AI; explain automated decisions in understandable language; ensure human oversight for high-impact decisions; allow override and appeal; document the system for audit. These are not just legal tasks — they are design patterns.
Even outside high risk, the AI Act requires transparency for generative AI (deepfakes, synthetic content) and chatbots — disclosure obligations the designer implements. Ignoring these requirements is not "creative grey zone" — it is non-compliance.
Useful frameworks beyond law: Google PAIR (People + AI Research), IF Design Patterns, Microsoft Responsible AI Standard, Design Justice Network principles. None is perfect, but they provide shared vocabulary and reusable patterns.
Create concise internal guidelines for your team: when to use AI badges, disclosure templates, approved opt-in patterns, forbidden anti-patterns (AI fake urgency, opaque pricing), escalation criteria ("if it involves health/credit/HR → mandatory review"). Update them when regulations change or you learn from incidents.
AI designer ethics does not end with a course — it evolves with technology and norms. Subscribe to newsletters (Ada Lovelace Institute, AI Now, Relatronica), join communities, share cases with colleagues. Professional solidarity reduces pressure to yield on problematic choices.