{"id":4028,"date":"2025-09-19T13:56:06","date_gmt":"2025-09-19T11:56:06","guid":{"rendered":"https:\/\/tuto.provacy.com\/?page_id=4028"},"modified":"2025-09-22T16:12:31","modified_gmt":"2025-09-22T14:12:31","slug":"trust","status":"publish","type":"page","link":"https:\/\/tuto.provacy.com\/en\/ais\/trust\/","title":{"rendered":"4. Trust"},"content":{"rendered":"

Trustworthy AI<\/h2>\n\n\n\n
\n\n<\/div>\n\n\n\n

The 7 criteria of trustworthy AI<\/strong> were proposed in 2019 by the High-Level Expert Group on Artificial Intelligence (HLEG)<\/strong>, mandated by the European Commission<\/strong>.<\/p>\n\n\n\n

\ud83d\udc49 Their goal: to define guidelines ensuring that AI is legally, ethically, and technically reliable<\/strong>.<\/p>\n\n\n\n

The 7 criteria<\/h3>\n\n\n\n
    \n
  1. Human oversight<\/strong>: AI must remain under human control, without uncontrolled autonomy.<\/li>\n\n\n\n
  2. Robustness and safety<\/strong>: resilience against errors, attacks, and failures.<\/li>\n\n\n\n
  3. Privacy and data governance<\/strong>: responsible and secure management of personal data.<\/li>\n\n\n\n
  4. Transparency<\/strong>: explainability of decisions, traceability of processes.<\/li>\n\n\n\n
  5. Diversity, non-discrimination, and fairness<\/strong>: avoiding bias and ensuring inclusiveness.<\/li>\n\n\n\n
  6. Societal and environmental well-being<\/strong>: contributing positively to society and minimizing ecological impact.<\/li>\n\n\n\n
  7. Accountability<\/strong>: existence of redress and audit mechanisms in case of issues.<\/li>\n<\/ol>\n\n\n\n

    \u2696\ufe0f These principles are not legal obligations like in the AI Act, but an ethical framework<\/strong> that strongly inspired the European regulation and the ALTAI<\/strong><\/a> checklist (a practical assessment tool).<\/p>\n\n\n\n

    Trust tab<\/h5>\n\n\n\n

    On the left, the form provides 6 to 10 questions per theme (55 in total). The scores, both overall and by theme, are calculated automatically as you go. On the right, targeted recommendations appear for each negative response.<\/p>\n\n\n\n

    \n
    \"\"<\/figure>\n<\/div>\n\n\n\n

    Some themes begin with a preliminary question that can automatically populate the questionnaire. For example: if the system does not process any personal data, then all the answers in the Privacy & data governance<\/em> theme are automatically set to Not applicable<\/em>.<\/p>\n\n\n\n

    \n
    \"\"<\/figure>\n<\/div>\n\n\n\n
    \n

    \ud83d\udca1From the Links tab, you can create activities \"\" related to the AI System form<\/p>\n<\/blockquote>\n\n\n\n

    \n
    <\/div>\n\n\n\n
    <\/div>\n\n\n\n
    <\/div>\n<\/div>\n\n\n\n<\/i><\/span>\n