CVE-2026-82579

AshAi invalid tool calls could repeatedly resend unchanged model requests

ash-project · ash_ai

First documented disclosure · Review checkpoint

Summary

Filtered-out tool calls could prevent conversation progress.

Severity and exploitation evidence

Medium · 6.0/10 · CVSS 4.0. Severity source.

Exploitation not established: Malicious exploitation is not established by the reviewed originals. Public demonstrations and catalog absence do not establish exploitation status.

Exploitation evidence source

Technical severity is not a measure of observed attacks or of risk to your deployment. Attempts, demonstrations, and successful compromise retain the qualifications in the source.

How the weakness works

An empty list after validating names and removing spent tool-call IDs recursed with unchanged history.

Conditions required

An attacker can influence model output, including through prompt injection, to produce invalid or previously processed calls.

Potential impact

Repeated model requests consume time and budget. Execution is unbounded only with max_iterations set to infinity; otherwise it exhausts the configured iteration budget.

Affected versions and configurations

ash_ai: 0.6.0 through versions before 1.0.0.

Fix and mitigation status

1.0.0.

Practical next steps

Apply the issue-specific correction: 1.0.0.

Operating lesson

Stop a tool loop when filtering leaves no actionable calls.

Evidence and limits

Earliest reviewed public evidence 2026-08-30; CVE publication 2026-08-31. CVSS 4.0 6 is attributed to EEF in the CNA record.

Sources