#NØØT_Security_Alerts 🚨 #ALERT — Network-AI A2A Server Authentication Bypass Exposes Agent Task Operations October 17, 2026 PRODUCT: Network-AI (network-ai npm) | CVE-2026-120314 | GHSA-qvf5-9prv-7r9c | HIGH (CVSS 3.1: 7.3). AFFECTED: 5.14.0–5.15.4 according to the VulnCheck CNA record. The project advisory lists versions through 5.15.4 without establishing the earliest affected release. FIXED: 5.15.5, released October 17. IMPACT: When A2AServer has no configured secret, it skips Bearer authentication. Anyone able to reach the endpoint can submit agent tasks, retrieve task results by ID, or cancel tasks without credentials. Submitted text reaches the application-defined executor; further impact depends on its capabilities. The listener defaults to 127.0.0.1. External exposure requires a reachable binding or another access path; it must not be assumed. EXPLOITATION STATUS: The project advisory documents successful unauthenticated HTTP task execution. No malicious in-the-wild exploitation is established. The official CVE was published November 2, after the original disclosure and fix. TRIAGE: Inventory A2A listeners, secret configuration and reachable endpoints. Examine available HTTP and application logs for unexpected tasks/send, tasks/get, and tasks/cancel requests. Logging gaps limit retrospective assurance. URGENT ACTION: Restrict untrusted access immediately. Upgrade to 5.15.5 or later, configure a nonempty secret, and verify unauthorized requests are rejected. Explicit unauthenticated mode is restricted to loopback and intended for local use. Account for the breaking configuration change, then investigate suspicious executor activity and credential exposure where warranted. SOURCE — PROJECT ADVISORY: https://github.com/Jovancoding/Network-AI/security/advisories/GHSA-qvf5-9prv-7r9c OFFICIAL CNA RECORD: https://github.com/CVEProject/cvelistV5/blob/main/cves/2026/120xxx/CVE-2026-120314.json VENDOR FIX AND RELEASE HISTORY: https://network-ai.org/documentation/changelog/ CONFIDENCE: HIGH for the documented bypass, prerequisites and fix. Actual deployment exposure, malicious exploitation and downstream compromise remain unverified. #CyberSecurity #AppSec #DevSecOps #IncidentResponse