Technical impact
The guardrails‑detectors component of Red Hat OpenShift AI exposes a public detection API that accepts a regular expression supplied in the detector_params.regex field. The flaw, a classic Regular Expression Denial of Service (ReDoS) classified as CWE‑1333, allows a remote attacker to craft a regex that triggers catastrophic backtracking. This results in an idle worker process consuming 100 % CPU for an extended period and causing a denial of service for the entire guardrails‑mediated LLM pipeline.
Affected Systems
This vulnerability affects deployments of Red Hat OpenShift AI that include the guardrails‑detectors component. All shipped versions that contain this component are potentially susceptible; no specific version boundaries are indicated in the available data.
Risk and Exploitability
The CVSS score of 6.5 indicates moderate severity, while the EPSS score of less than 1 % suggests low likelihood of widespread exploitation. The vulnerability is remote and unauthenticated because the API is publicly accessible without credentials, and it is not listed in the CISA KEV catalog. The inference that the attack vector is remote and unauthenticated comes from the public accessibility of the API, which is described as publicly accessible without credentials. If successfully exploited, the denial of service would degrade the availability of the LLM pipeline for all users of the affected OpenShift AI instance.
