Detecting LLMjacking: validate, enumerate, invoke
Cost anomalies cannot tell a busy week from a hijacked key. The kill chain can: first-ever credential validation, model enumeration, then invocation.
2026-09-29
How the detections are built
Working notes rather than announcements: the design decisions behind the rules, the ones that were wrong first, and the identifiers you can go and read in the source.
Correlating CI tampering: the pair is the finding
A changed workflow file is an ordinary Tuesday. A changed workflow in a repository that holds a deploy credential is a finding. How the join works.
2026-10-06
Proving a finding six months later
Detection is the easy half. What has to be true at detection time for a finding to still be defensible in a review six months later.
2026-09-15
Why the detection path has no LLM in it
A model in the scoring path costs you reproducibility, an explainable queue and regression tests. What it is genuinely good for, and where the line is.
2026-09-22