Aix-DB CVE-2026-8335

Attacker must have network reachability to the Aix-DB HTTP service from an adjacent network position (CVSS AV:A) - typically the same L2 segment, VPN, or routed internal subnet as the deployment, not arbitrary internet origin. The target must be running Aix-DB at or below version 1.2.4 with the /llm/process_llm_out endpoint exposed; no authentication, user interaction, or non-default configuration is required because the endpoint itself omits the token validation enforced on other routes. Impact is limited to data the application's database user can read via SELECT - write, delete, or DDL operations are not implicated, and stored procedures or non-SELECT statements may not execute depending on the endpoint's SQL handling.

The CVSS 4.0 base score of 7.1 is driven by AV:A (adjacent network) rather than AV:N, meaning the attacker must share a broadcast or routable segment with the Aix-DB host - for many internal deployments this is functionally equivalent to network-reachable, but it does limit drive-by internet exploitation. Confidentiality impact is High (VC:H) while integrity and availability are Not Impacted, consistent with a SELECT-only data read primitive. There is no EPSS score, KEV listing, or public POC referenced in the supplied intelligence, so exploitation is plausible-but-not-observed; the gating factor is real-world exposure of Aix-DB instances on reachable subnets, not technical difficulty. CERT-PL reporting lends credibility to the technical claim even though independent vendor confirmation is absent.

An attacker who has gained a foothold on the same adjacent network as an Aix-DB deployment - for example via a compromised workstation, rogue Wi-Fi client, or lateral movement from another internal service - sends a crafted HTTP POST to /llm/process_llm_out carrying a SELECT statement against sensitive tables. Because the endpoint never invokes the token check used by sibling routes, the query executes and the response leaks rows directly to the attacker, who can iteratively enumerate schemas and extract data such as credentials, user records, or LLM prompt history. No public exploit identified at time of analysis, but the CERT-PL writeup describes the missing check at a level of detail that makes scripted abuse straightforward.