AI Model Watermarking and Provenance
Watermarking and provenance are the two most confused terms in AI security. A practical breakdown of what each actually does, where the 2025 techniques break, and what to ship in the meantime.
Deep dives, practical guides, and incident analyses from engineers who build Safeguard. No fluff, no vendor FUD — just what you need to ship secure software.
Watermarking and provenance are the two most confused terms in AI security. A practical breakdown of what each actually does, where the 2025 techniques break, and what to ship in the meantime.
SLSA v1.2 was approved in November 2025 and finally completes the Source Track that v0.1 only sketched. We break down the new source levels and what producers must change.
Two and a half years after npm provenance launched, adoption is climbing but uneven. Here is the late-2025 picture across the top packages and frameworks.
Provenance answers where software came from and how it was built. Here is how to implement end-to-end provenance tracking from source to deployment.
SLSA v1.1 was approved in April 2025 with the Build track stabilized. We dig into the spec changes, what L2 and L3 verifiers must reject, and how producers should re-evaluate provenance.
Generate and validate SLSA v1.0 provenance attestations in GitHub Actions using slsa-verifier, gate releases on builder identity, and prove build integrity.
Trademarks matter in open source security because they are the signal of authentic origin. When trademark policies fail, typosquatting, impostor forks, and compromised builds follow.
A step-by-step tutorial for publishing npm packages with provenance attestations so your consumers can cryptographically verify the build source.
The SLSA framework reached v1.0 in April 2023, providing a practical framework for software supply chain integrity that's already being adopted by major package registries.
Weekly insights on software supply chain security, delivered to your inbox.