Autodesk Fusion CVE-2026-0535

Exploitation requires that an attacker can place a malicious HTML payload into the description field of a Fusion component AND that a victim user opens and clicks that component's description within the Fusion desktop app - this maps directly to the CVSS UI:R requirement. The attacker does not need credentials on the victim's machine (PR:N), but does need a delivery channel where the victim will encounter the poisoned component (shared team project, cloud-hosted library, third-party part supplier, or imported file). Exploitation is limited to users running the unpatched Autodesk Fusion desktop client on Windows or macOS; users of read-only viewers or web-only Fusion workflows that do not render component descriptions through the desktop rendering layer are not in scope.

Signals are mixed and lean toward moderate priority rather than emergency. CVSS 8.1 is driven by network attack vector, low complexity, no privileges required, high confidentiality and integrity impact, but is gated by required user interaction (UI:R) and unchanged scope. Counterbalancing this, EPSS is 0.02% (5th percentile) and there is no KEV listing or public POC, so widespread automated exploitation is unlikely. The 'AV:N' rating warrants scrutiny - exploitation actually requires a victim to open a malicious Fusion component and click a crafted description, which is more akin to a phishing-style social-engineering chain than a typical network-reachable RCE, so real-world risk is concentrated in environments where users routinely import third-party Fusion components or shared cloud projects. Tags list both XSS and RCE, which is consistent with the description's claim that script execution in the desktop context can reach local code execution.

An attacker uploads or shares a Fusion component (for example through a community library, shared team project, or supplier-provided model) whose description contains an HTML/JavaScript payload. When an engineer opens the project and clicks the component to view its description, the script executes inside the Fusion process and exfiltrates local files (project data, credentials, design IP) or drops and runs additional code as the current user. No public exploit identified at time of analysis.