We are writing a small number of longer papers on problems that do not fit in a blog post. None are published yet. The titles below are planned; subscribe and we will email you when each one is released.

Planned titles — coming soon

Title What it will cover
Governing AI agents: least privilege, approvals and containment Why agents need different controls from models; tool manifests, default-deny, approval thresholds and kill-switch design (EI-07, IS-11, IS-13, IS-15)
Evidence, not assertions: designing an AI assurance programme Evidence labels, control testing, tamper-evident logs and measuring time-to-produce evidence (SE-01, SE-05, SE-08)
Risk tiering that people trust Building tier criteria from risk appetite; calibrating triage; human confirmation (AO-04, EI-02, GL-01)
One control set, many frameworks How to use indicative mappings across NIST AI RMF, ISO/IEC 42001 and the EU AI Act without overclaiming
The inventory problem: finding and owning enterprise AI Discovery, reconciliation and attestation for models, agents and embedded vendor AI (EI-01, EI-08, EI-09, EI-11)

VERITRAIL controls cited

Framework mappings are indicative VERITRAIL interpretations (Control Library v1.0.0); they are not legal advice and do not establish conformity or compliance.

In the meantime

Use the form on this page to subscribe. We send no more than one or two emails a month and you can unsubscribe at any time. See the privacy notice.

Subscribe

Last reviewed: