Two questions we keep getting this month from teams prepping late-summer submissions:
1. 'Does a PCCP cover cybersecurity changes?' Short answer: only the cyber changes you explicitly scope into the modification protocol. Retraining an AI/ML model on new data is a PCCP-covered modification - the resulting shift in adversarial-robustness posture is a 524B cybersecurity concern that must be re-evaluated under your SPDF.
2. 'Where does this show up in the submission?' The Feb 3, 2026 premarket cybersecurity guidance expects: (a) a threat model that names the ML surface, (b) a monitoring plan that detects post-deployment drift and adversarial signals, and (c) a link between your PCCP modification protocol and the cyber activities that re-run on each update (SBOM refresh, VEX re-evaluation, adversarial re-testing).
If you filed a PCCP without those cross-references, plan to amend before your next Q-Sub - reviewers are increasingly flagging the gap as a 524B deficiency, not a PCCP one.
- MDCG 2019-16 Rev.2 at six months: Notified Body audit patternsSix months in, Notified Body audits under MDCG 2019-16 Rev.2 show three clear failure modes - narrative-only SBOMs, CVD policies without operational evidence, and traceability gaps between security risk controls and design outputs.
- Three years of Section 524B: the deficiencies that never went awayThree-year retrospective on Section 524B enforcement: SBOM depth, VEX justification, and CVD operational evidence remain the top three categories of cyber deficiencies. Two new patterns emerged in 2026 - AI/ML threat-model gaps and weak end-of-support labeling.
- IEC 81001-5-1 Amendment 1 reaches FDIS - what MedTech teams should trackAmendment 1 to IEC 81001-5-1 (health software security lifecycle) is in Final Draft International Standard ballot. It sharpens SBOM, CVD, and postmarket vulnerability expectations - and it's the standard EU Notified Bodies are quietly aligning MDCG 2019-16 audits against.