Oligo Security disclosed two JFrog Artifactory flaws on Aug. 20 that let anonymous or low-privileged users poison package metadata without changing the artifact bytes. JFrog has issued fixes for CVE-2026-69106 and CVE-2026-65922.
The break is in the trust layer. One issue lets attacker-controlled URL data shape cached metadata, and the other lets writes reach trusted .jfrog/ paths that package handlers use for things like signing keys, OCI referrers, Docker indexes, and Ansible indexes. That means integrity checks on the stored package can still look clean while downstream tools are steered toward untrusted references.
For shared or internet-facing Artifactory instances, the exposure is not just a single repo entry but the metadata consumers use to decide what is legitimate. If build systems rely on that repository, a weak account can influence what other users and pipelines resolve even after the artifact itself remains untouched.