IntelliJ IDEA CVE-2026-86502

Exploitation requires that the IntelliJ IDEA Remote Development feature is actively in use, causing the IJent gRPC agent to be deployed and listening on the remote host. The attacker must have local access to that remote host - the CVSS AV:L/PR:N combination confirms local presence is required but no privileges are needed once local access is established. Standard IntelliJ IDEA usage without the Remote Development feature does not expose the IJent gRPC server. Remote network attackers cannot reach the service unless the gRPC port is deliberately exposed externally, which is not the default. The vulnerability is most dangerous on shared or multi-tenant hosts; single-user dedicated developer machines represent substantially lower practical risk.

The CVSS 3.1 score of 8.4 (AV:L/AC:L/PR:N/UI:N/S:U/C:H/I:H/A:H) accurately characterizes this vulnerability: local attack vector with no privileges required and no user interaction, yielding full confidentiality, integrity, and availability impact. The local attack vector (AV:L) substantially limits the attack surface relative to network-reachable flaws, but PR:N confirms that no elevated privileges are needed - any process co-located on the remote host can exploit this. The risk is elevated in shared infrastructure scenarios such as multi-user cloud VMs, CI/CD runners, or developer jump hosts where multiple principals have local access. EPSS data is not present in the available intelligence; without it, exploitation probability cannot be quantified beyond the CVSS signal. No public exploit code has been identified and no CISA KEV listing exists. The practical risk to organizations scales directly with how broadly Remote Development is deployed on shared or multi-tenant hosts versus dedicated single-user machines.

An attacker with local access to a shared remote development host - such as another tenant on a multi-user cloud VM or a process running in a compromised CI/CD pipeline - locates the IJent gRPC server's listener port and connects using a standard gRPC client over plaintext without supplying any credentials. The attacker sends a crafted gRPC request directing the IJent agent to execute an arbitrary OS command, achieving code execution in the context of the developer's session or the IJent process itself. No public exploit code has been identified, but exploitation requires no specialized tooling beyond a gRPC client library.