Q1 · AI Act
Is our service agent high-risk?
Usually not by itself — but transparency duties apply from August 2026, and AI inside safety functions of machinery moves into high-risk scope by August 2028.
Four EU regulations now touch AI on machines. We turn them into autonomy levels, approval gates and audit trails — so compliance speeds agents up instead of freezing them.
Q1 · AI Act
Usually not by itself — but transparency duties apply from August 2026, and AI inside safety functions of machinery moves into high-risk scope by August 2028.
Q2 · Machinery Regulation
From 20 January 2027, safety components with self-evolving ML behaviour require notified-body assessment. Agents that act on machines need clear boundaries to stay outside safety functions.
Q3 · Data Act
Users can now access — and share — the data your machines generate. Third-party service gets easier. The best defence is a service experience competitors can't match.
Level L2 · Acts with approval
Executes the fix on the machine after a named engineer signs off.
Controls required
Inventory every AI use — shadow AI included — classify it under the AI Act and Machinery Regulation, and leave with a prioritised gap list.
2 weeksAutonomy levels, action permissions, approval gates, audit trails and a kill switch — designed for agents that touch real systems.
4–6 weeksPolicies, roles, risk and impact assessments and a lifecycle process that can grow into certification.
90 daysRole-specific training for engineers, service teams and managers — documented to show the measures Art. 4 asks for.
OngoingWe are engineers, not lawyers. Our work turns regulation into technical and organisational controls, and we coordinate with your legal counsel and notified bodies where required.
Release for production
A 30-minute call is enough to find the first agent worth building — and the controls it will need.