Responsible AI
AI should strengthen human judgement, with clear ownership of data, risk and decisions.
This is a public review draft. It is not an offer or an operative agreement; signed service terms and approved schedules prevail.
Inventory and purpose
Before an AI use case is adopted, identify its intended purpose, operator, data sources, affected people and decision impact. Assess the applicable AI Act role and risk classification for the concrete system rather than treating every tool as equivalent.
Human oversight and assurance
Set review points for material outputs, test accuracy and failure modes, retain traceability to sources and make escalation possible. Do not put client-confidential information into an AI service without authority, security review and appropriate terms.
Transparency and skills
Provide disclosures when an applicable transparency rule requires them. Support AI literacy for staff and other persons operating AI on our behalf, taking account of their role and context under the amended Article 4. The design should be revisited as the law and system change.