Impact
The defect lies in the IsValid function of Ollama’s Pull API, which mis‑handles model names. Based on the description, a remote attacker can craft an API call that causes the server to perform arbitrary outbound requests. This server‑side request forgery can be used to exfiltrate data, reach internal services, or be a stepping stone for further attacks.
Affected Systems
The vulnerability affects the Ollama community AI tool. All releases up to and including version 0.34.1 are vulnerable. The fix is delivered in version 0.34.2, which incorporates the commit identified by dfabde4539e42ba1e1eab50a3a50b88aea7958a0.
Risk and Exploitability
The CVSS score of 6.9 indicates moderate severity. EPSS data is not available and the issue is not listed in the CISA KEV catalog. Attackers can trigger the vulnerability remotely by sending a forged Pull request; the server is then compelled to contact arbitrary URLs, potentially exposing internal assets. The exploitation requires only the ability to reach the exposed Pull API endpoint.
OpenCVE Enrichment