The EU Council of Ministers adopted the simplified AI Act Omnibus package on June 29. The decision extends high-risk deadlines to December 2027 (Article 6) and August 2028 (Annex III), brings the ban on nudification apps into force from December 2026, and grants SMEs an extended grace period. Publication in the Official Journal is expected within days, making the regulation formally binding.
Novudom Compliance Briefing W.27 2026
This week marks two milestones: the EU Council's final approval of the AI Act Omnibus and the entry into force of Sweden's incident reporting rules under the Cybersecurity Act. Meanwhile, the Novo Nordisk data extortion shows that cyber threats now target organisations' most valuable intellectual property directly, and DORA supervision transitions from document review to automated real-time verification.
AI Act Omnibus: EU Council gives final approval
The timeline for high-risk AI systems has been extended, but prohibition provisions (Article 5) tighten faster. Review your AI inventory now: systems classified as prohibited must be decommissioned by December 2026. High-risk compliance has more time, but not unlimited.
Cybersecurity Act: Incident reporting rules take effect July 1
On July 1, 2026, the new incident reporting rules under Sweden's Cybersecurity Act took effect. Essential and important entities must now report significant incidents to NCSC within 24 hours (early warning), followed by a full report within 72 hours and a final report within one month. The rules follow NIS2 Directive Article 23 and apply to all 18 sectors.
If you are classified as essential or important, reporting requirements are now in force. Ensure your incident response process has updated contact channels to NCSC, that the 24/72 hours/1 month deadline is built into procedures, and that a reporting officer with deputy is designated.
Novo Nordisk: Cyber extortion targeting clinical data and AI models
Novo Nordisk confirmed on June 11 that 1.3 TB of data was stolen in a targeted cyber attack. The attackers demand $25 million to not publish clinical trial data and an additional $50 million for stolen AI models used in drug development. The breach exposes patient data, research results, and proprietary AI.
Cyber extortion is no longer just about downtime. Clinical data, AI models, and research results are primary targets. Classification of information assets and segmentation of access to sensitive datasets should be a board agenda item for Q3.
ENISA Cyber Europe 2026: Transport sector cyber resilience tested
ENISA conducted the eighth edition of Cyber Europe on June 10-11, the EU's largest cross-sector cyber exercise. This year tested rail and maritime network resilience, focusing on incident coordination between national CSIRTs, operators, and supervisory authorities. Over 1,200 participants from 30 countries took part.
Transport is one of the sectors ENISA assesses as having the lowest NIS2 maturity. If you supply to or collaborate with the transport sector, expect tightened cybersecurity requirements in contracts and procurement. Supply chain requirements cascade down the chain.
IQVIA fined EUR 5M by CNIL for health data handling failures
French data protection authority CNIL issued a EUR 5 million fine against healthcare analytics company IQVIA on May 26. The fine concerns inadequate handling of a health data warehouse where pseudonymised patient data was processed without sufficient security measures. CNIL found that IQVIA had not conducted adequate DPIAs and that data access was insufficiently restricted.
Health data is the data category generating the largest GDPR fines in 2026. CNIL's decision confirms that pseudonymisation does not replace other security measures: DPIAs are still required, and access controls must be proportionate and documented. If you process sensitive personal data (Article 9), verify that your DPIAs are current.
Ford Mexico: Krybit ransomware threatens data leak
Ransomware group Krybit claimed on June 28 to have stolen data from Ford Mexico's internal systems, including customer credentials from SSO and login portals. The group threatens publication unless ransom is paid. The case illustrates how regional subsidiaries and local IT environments often have weaker protection than the parent company's central infrastructure.
International groups with regional subsidiaries should assess whether local IT environments meet the same security standard as central infrastructure. SSO and authentication portals are primary attack surfaces. If you operate across countries: does the local IT manager or the group CISO own incident response?
DORA: Automated supervision and real-time ICT register verification
European Supervisory Authorities (ESAs) have begun using automated tools to cross-reference financial institutions' ICT registers (Register of Information) against operational data. Supervision is shifting from document review to real-time verification, comparing registers against actual vendor usage, incident history, and contract terms. Institutions that reported incomplete registers have already received orders.
Your ICT register is no longer just a compliance document but a living supervisory record subject to automated review. Ensure the register matches actual vendor usage, that incident history is linked to the correct provider, and that updates are continuous.
Weekly reflection
Three of this week's seven articles point to the same fundamental problem: unclear ownership. Novo Nordisk's AI models lacked sufficient segmentation. Ford's regional IT environments fell outside the group's central control. Financial institutions' ICT registers did not match reality. Compliance is not about documents that pass an audit, but about structures that work under pressure.