JetBrains TeamCity Deserialization of Untrusted Data Vulnerability

Technical Analysis

CVE-2026-63077 arises from unsafe deserialization of untrusted data transmitted through TeamCity's build agent polling protocol, allowing an unauthenticated attacker to craft malicious serialized payloads that execute arbitrary code on the TeamCity server or connected build agents. Because TeamCity servers typically hold sensitive credentials, source code access, and CI/CD pipeline secrets, successful exploitation can lead to full build-chain compromise and downstream software supply-chain attacks. The short CISA remediation deadline (three days) strongly suggests confirmed in-the-wild exploitation or a public PoC. TeamCity build agents are frequently used to automate deployment of AI/ML pipelines, model training jobs, and agent framework dependencies (e.g., pulling packages for LLM tool integrations or RAG data pipelines), so compromise of the build server could allow attackers to inject malicious code or credentials into artifacts consumed by downstream AI agent systems. Any organization using TeamCity to build, package, or deploy AI agent software should consider this a direct supply-chain risk to their agent infrastructure.

Affected Systems

JetBrains TeamCity on-premises server installations exposed to network access from build agents, particularly configurations where the agent polling protocol endpoint is reachable by untrusted or external hosts; specific vulnerable version ranges should be confirmed against JetBrains' official security advisory once published.

Indicators of Compromise

• No public IOCs (hashes, IPs, domains) available at time of publication; monitor CISA KEV catalog and JetBrains security advisories for updates.

Remediation Steps

1. 1

   Apply vendor patch immediately

   Upgrade TeamCity to the patched version specified in JetBrains' security bulletin as soon as it is released; treat as emergency out-of-band patching given the CISA KEV due date.

2. 2

   Restrict network exposure

   Limit access to the TeamCity server and agent polling endpoints to trusted internal networks only; block external/internet exposure via firewall rules.

3. 3

   Rotate credentials and secrets

   Rotate all API keys, service account credentials, and secrets stored in or accessible from TeamCity, including any credentials used by downstream CI/CD or AI agent deployment pipelines.

4. 4

   Audit build artifacts

   Review recent build outputs and deployed artifacts for signs of tampering, especially those feeding AI agent, RAG, or LLM tool deployments.

5. 5

   Enable monitoring and logging

   Increase logging on the TeamCity server and agents to detect anomalous deserialization attempts or unexpected process execution.