The name, not the licence — the CRA reporting clock started on Friday, and not for everyone
Article 14 of the Cyber Resilience Act, Regulation (EU) 2024/2847, became applicable on 11 September 2026 and ENISA's Single Reporting Platform went live the same day: an early warning within 24 hours of awareness, a notification within 72, a final report within 14 days of a corrective measure, to the CSIRT designated as coordinator and ENISA rather than to a data protection authority, and triggered by reliable evidence that a malicious actor exploited the vulnerability in a system — not necessarily yours. The obvious reading, that the whole Regulation is a 2027 problem, is right about the main obligations and wrong about this one. The second and larger correction is that it did not start for everyone: Article 71(2) puts open-source software stewards on Article 24(3) from 11 December 2027, while Article 3(13) makes anyone who markets a product with digital elements under its own name or trademark a manufacturer, free of charge or not — so the licence does not decide the clock, the name does. The exposure inverts in the same direction: Article 64(2) puts Article 14 in the top penalty tier, EUR 15 million or 2.5% of worldwide turnover, alongside the essential requirements, while Article 64(10) exempts stewards from administrative fines entirely and exempts micro and small enterprises from fines for the 24-hour early warning specifically. The practitioner moves are to label every component in the agent stack manufacturer, steward or neither; to write the runbook trigger as evidence of exploitation anywhere rather than as an alert in your own estate; and to keep the 24-hour CRA clock separate from GDPR Article 33's 72 hours, because they run to different authorities.
Post
Yesterday a web form on ENISA's servers became mandatory. Most engineers I know still file the Cyber Resilience Act under "2027 problem".
Article 14 started on 11 September 2026. Regulation (EU) 2024/2847 applies in full from 11 December 2027 — the reporting duty did not wait for it, and ENISA's Single Reporting Platform went live the same day.
What actually changed for anyone shipping an agent stack in the EU:
-
The trigger is not your incident. Article 3(42) defines an actively exploited vulnerability as one with reliable evidence that a malicious actor has exploited it in a system, without the system owner's permission. Not your system. A system. Your SIEM can be silent and the clock still runs.
-
24 hours for the early warning, 72 for the notification, 14 days for the final report once a fix exists. Severe incidents: 24, 72, one month. GDPR gives you 72 hours for a personal data breach — this runs three times faster, and it goes to your CSIRT coordinator and ENISA rather than to a data protection authority.
-
It did not start for everyone. Article 71(2) puts open-source software stewards on Article 24(3) from 11 December 2027. Manufacturers started Friday.
-
Article 3(13) makes you the manufacturer if you market the product under your own name or trademark. Free of charge does not exempt you. The licence does not decide this — the name does.
-
The exposure inverts. Article 64(2) puts Article 14 in the top penalty tier, EUR 15 million or 2.5% of worldwide turnover, whichever is higher, alongside the essential requirements. Article 64(10) says stewards cannot be fined at all, and micro and small enterprises cannot be fined for missing the 24-hour early warning specifically — every other duty still stands for them.
For scale on what you would be reporting: Meta published Muse's security architecture on 8 September and prices a successful single-user prompt injection at up to $130,000 of a $300,000 maximum. That is the industry valuing the failure class that now opens a 24-hour window.
My Monday: one column beside every component in the stack — manufacturer, steward, or neither. An afternoon, and it decides who files.
Which of your components has your name on it?
#CyberResilienceAct #AIAct #AgenticAI #EUTech #DevSecOps
First comment
Sources, in the order the argument uses them.
ENISA's FAQ for the Single Reporting Platform — the 11 September 2026 start, and the Article 71(2) split that puts stewards on 11 December 2027: https://www.enisa.europa.eu/topics/product-security/single-reporting-platform-srp/frequently-asked-questions
The Commission's CRA page — entry into force 10 December 2024, reporting from 11 September 2026, main obligations from 11 December 2027: https://digital-strategy.ec.europa.eu/en/policies/cyber-resilience-act
The Commission on open source — quotes Article 64(10) directly: stewards are not subject to administrative fines: https://digital-strategy.ec.europa.eu/en/policies/cra-open-source
Article 14 itself — the 24/72/14-day and 24/72/one-month sequences, and who receives them: https://www.european-cyber-resilience-act.com/Cyber_Resilience_Act_Article_14.html
Article 64 — the three penalty tiers, and the microenterprise carve-out for the 24-hour deadlines only: https://www.european-cyber-resilience-act.com/Cyber_Resilience_Act_Article_64.html
Meta's Muse security write-up, 8 September 2026 — the architecture, and the bounty figures: https://research.meta.ai/blog/security-and-safety-for-ai-agents-our-approach-with-muse
The Register's coverage of the start date, 11 September 2026: https://www.theregister.com/security/2026/09/11/eus-cyber-resilience-act-starts-the-24-hour-vulnerability-clock/5295821
Cut for length: the Commission published practical CRA guidance on 27 July 2026, which is the thing to read before you write the runbook rather than after.
Claims
- CLAIM-2026-09-12-01
- CLAIM-2026-09-12-02
- CLAIM-2026-09-12-03
- CLAIM-2026-09-12-04
- CLAIM-2026-09-12-05
- CLAIM-2026-09-12-06
- CLAIM-2026-09-12-07
- CLAIM-2026-09-12-08
- CLAIM-2026-09-12-09
Sources
- europeancyb2024definitio
- theregister2026eus
- enisa2026faq
- intersoftco2026article
- europeancom2026opensourc
- europeancyb2024penalties
- europeancyb2024reporting
- metaairesea2026security
- europeancom2026timeline
Corrections
None.
AI disclosure
The video for this edition contains AI-generated media: the narration is a synthetic voice, the cover art and any background imagery are generated, and the brand ident animation is generated. The script, the claims and the source verification are human-authored and human-checked. Assets carry IPTC DigitalSourceType and C2PA Content Credentials.