Berlin — The German federal government has issued an emergency regulation requiring operators of artificial‑intelligence systems that affect public safety to report incidents to a central AI Alert Centre within 24 hours, the Federal Ministry for Digital Affairs said on Monday, 21 September 2026. The rule, signed by State Secretary Lukas Vossberg, applies to systems used in transport, energy supply, healthcare and emergency services and is effective immediately.
Under the order, designated 'AI Incident Notification Framework v1.0', firms must register significant malfunctions, outages or behaviour that posed or could have posed a risk to life, health or public order with the new alert hub run by the Federal Office of Civil Protection and Disaster Assistance (BBK). Reports must include a timeline, system logs where feasible, and the operator's initial mitigation steps. Failure to report can carry administrative fines of up to €5 million (about $5.4 million) or 2% of annual worldwide turnover, whichever is higher.
Vossberg said the measure was designed to speed up national situational awareness and co‑ordinate responses. "We have seen that delays in recognising and sharing information about AI failures compound risks to citizens," he said. "This framework gives emergency services and operators the shared visibility they need while preserving data‑protection safeguards." The ministry said technical guidance for the format and transmission of alerts would be published within seven days.
The announcement follows a string of locally significant incidents this year that exposed gaps in oversight of automated decision systems. In July a fatal fall at a Neumarkt housing estate in Saxony — where an automated door‑control routine prevented first responders from entering a stairwell quickly — and persistent scheduling errors on the Greenline tram project in the Rhine‑Main region were cited in ministry papers as illustrative cases that would have benefited from faster cross‑sector alerts. Ivan Esteves‑Melo, a transport union representative, said the new rule was overdue. "Operators and crews need timely information. When a scheduling system fails and no one outside the company knows, passengers and staff are put at risk," he said.
Industry groups warned the requirement will add operational and legal burdens for smaller firms and start‑ups. Mack Lowe, speaking for the Federation of Industrial AI Providers (a new umbrella trade body), said compliance will require additional audits and insurance costs, and urged the government to provide technical and financial support for compliance. By contrast, Mathilde A. Fontaine of Consumers for Digital Rights welcomed the move. "Transparency about failures that affect safety is the minimum citizens should expect," Fontaine said. "Timely public reporting can prevent harm and restore trust."
The government said the emergency regulation was enacted under powers in the Digital Security Act and will be debated in the Bundestag within 60 days, where lawmakers may amend it or embed it into permanent law. Data‑protection experts expect discussions to focus on how to share diagnostic logs without exposing personal data or trade secrets. Kornél Szabó, head of IT at a university hospital in Frankfurt required to register under the rule, described the ministry's guidance as "practical but compact" and said his institution had already begun mapping which systems fall under the obligation.
Practical next steps for affected operators include mandatory registration with the BBK portal within 48 hours, submission of a contact and incident liaison officer, and a first quarterly public summary of incidents affecting safety starting 1 January 2027. The ministry said it will open bilateral talks with EU partners to align reporting standards and expects an initial coordination paper at an EU digital ministers' meeting in December. "AI knows no border; our response must be coordinated across Europe," Vossberg added. The BBK said its hotline for operator registration would go live on Tuesday morning.
