CVSS 4.0
5.9 mediumCVSS:4.0/AV:N/AC:L/AT:P/PR:L/UI:P/VC:N/VI:H/VA:N/SC:N/SI:N/SA:NMedium (5.9) in cohere-ai cohere-python. EPSS 0.3% estimated probability of exploitation activity in the next 30 days.
CVSS:4.0/AV:N/AC:L/AT:P/PR:L/UI:P/VC:N/VI:H/VA:N/SC:N/SI:N/SA:NCVSS:3.1/AV:N/AC:H/PR:L/UI:R/S:U/C:N/I:H/A:NExplain CVE-2026-108597 for an engineer deciding whether to act. Use https://cve.bz/api/public/cves/CVE-2026-108597?via=ask-ai-copy and https://cve.bz/cve/CVE-2026-108597?via=ask-ai-copy as the source of truth: what is affected and which versions fix it, whether it is in CISA KEV, public exploits, and its EPSS (an internet-wide 30-day estimate of exploitation activity, not a forecast for my systems). If they give no fixed versions, use the CVE record https://cveawg.mitre.org/api/cve/CVE-2026-108597 or the vendor advisory, not third-party mirrors. Tell me how to check whether I am affected and how to verify a fix. Cite the sources and say where they are uncertain.Opens a third-party assistant; its answer is its own, not cve.bz's. The prompt holds only this CVE's id and public links.
From the CNA's record in the CVE List.
| Product | Affected |
|---|---|
| cohere-ai cohere-python | from 5.11.0 through 7.2.0 |
Red Hat rates this Moderate.
A MODERATE severity path traversal (tar slip) flaw was found in the Cohere Python SDK. The SDK's AWS/SageMaker helper code downloads a model tar archive from an Amazon S3 bucket and extracts it using tarfile.extractall without validating member file paths. An attacker who has write permissions to the S3 bucket or prefix from which the victim downloads model archives can supply a specially crafted archive containing absolute paths or '../' directory traversal sequences, causing arbitrary files on the host running the SDK to be overwritten when the archive is extracted. The impact is limited to integrity (arbitrary file overwrite with no direct confidentiality or availability impact), and exploitation requires both attacker write access to the associated S3 bucket and a user or automated process to trigger the download-and-extract operation. This is reflected in the Red Hat CVSS 3.1 score of 4.8 (CVSS:3.1/AV:N/AC:H/PR:L/UI:R/S:U/C:N/I:H/A:N), which matches the upstream CVSS 3.1 score, so there is no meaningful severity difference.
Red Hat OpenShift AI (rhoai-3.4 and rhoai-3.5 streams, via the odh-trustyai-garak-lls-provider-dsp-rhel9 container image) and Lightspeed Core (lightspeed-core-0.6, via lightspeed-stack-rhel9) bundle affected versions of the cohere Python package and are considered affected. The resolution for these products is DEFER: because the flaw is rated MODERATE with a CVSS score below 7.0, the fix is deferred to a future product update rather than remediated immediately. The residual risk is considered limited given the high attack complexity and the attacker's requirement for pre-existing write access to the S3 bucket serving the model archives.
The attack requires an attacker to be able to write a model archive to the Amazon S3 bucket (or prefix) from which the Cohere SDK downloads and extracts archives. As a workaround, restrict write access to that S3 bucket/prefix to trusted identities only, for example by reviewing IAM and S3 bucket policies and removing write permissions from untrusted principals, and avoid invoking the SDK's SageMaker model download-and-extract flow against buckets whose contents are not fully trusted.
| Product | Package | State | Advisory |
|---|---|---|---|
| Lightspeed Core | lightspeed-core/lightspeed-stack-rhel9 | Fix deferred | |
| Red Hat OpenShift AI (RHOAI) | rhoai/odh-trustyai-garak-lls-provider-dsp-rhel9 | Fix deferred |
Contains Red Hat Security Data © Red Hat, Inc., licensed under CC BY 4.0. Source: https://access.redhat.com/security/data. Modified: reformatted. Red Hat's page for CVE-2026-108597.
The attack patterns MITRE's CAPEC relates to this record's weaknesses, as MITRE names them. Credits.