CVE-2026-28277
Published: 05 March 2026
Summary
CVE-2026-28277 is a medium-severity Deserialization of Untrusted Data (CWE-502) vulnerability in Langchain Langgraph. Its CVSS base score is 6.8 (Medium).
Operationally, exploitation aligns with the MITRE ATT&CK technique Exploitation of Remote Services (T1210); ranked in the top 43.5% of CVEs by exploit likelihood; it is not currently listed in the CISA KEV catalog.
This vulnerability is AI-related — categorised as AI Agent Protocols and Integrations; in the Data-Related Vulnerabilities risk domain.
The strongest mitigations our analysis identified are NIST 800-53 SI-10 (Information Input Validation) and SC-28 (Protection of Information at Rest).
Deeper analysis
CVE-2026-28277 affects LangGraph SQLite Checkpoint, an implementation of LangGraph CheckpointSaver that uses SQLite databases (both synchronously and asynchronously via aiosqlite), in versions 1.0.9 and prior. The vulnerability stems from the checkpointers' ability to load msgpack-encoded checkpoints, which reconstruct Python objects during deserialization. This can lead to unsafe object reconstruction if the checkpoint data is malformed.
Exploitation requires an attacker with privileged write access to the persistence layer, such as after compromising the SQLite database or gaining other high-privilege access. The CVSS v3.1 base score of 6.8 (AV:A/AC:L/PR:H/UI:N/S:U/C:H/I:H/A:H) reflects an adjacent network attack vector with low complexity but high privileges needed, resulting in high impacts to confidentiality, integrity, and availability. A crafted payload in the modified checkpoint data can trigger arbitrary object reconstruction upon loading, potentially enabling code execution or other severe effects (CWE-502: Deserialization of Untrusted Data).
The GitHub security advisory at https://github.com/langchain-ai/langgraph/security/advisories/GHSA-g48c-2wqr-h844 provides details on the issue, but no public patch is known. Practitioners should isolate checkpoint storage, validate inputs strictly, and monitor for unauthorized modifications to the backing database until a fix is available.
LangGraph is part of the LangChain AI ecosystem, making this relevant to AI/ML workflows relying on stateful graph persistence. No real-world exploitation has been reported.
EU & UK References
- 🇪🇺 ENISA EUVD: EUVD-2026-9860
Vulnerability details
LangGraph SQLite Checkpoint is an implementation of LangGraph CheckpointSaver that uses SQLite DB (both sync and async, via aiosqlite). In version 1.0.9 and prior, LangGraph checkpointers can load msgpack-encoded checkpoints that reconstruct Python objects during deserialization. If an attacker can…
more
modify checkpoint data in the backing store (for example, after a database compromise or other privileged write access to the persistence layer), they can potentially supply a crafted payload that triggers unsafe object reconstruction when the checkpoint is loaded. No known patch is public.
- CWE(s)
AI Security AnalysisAI
- AI Category
- AI Agent Protocols and Integrations
- Risk Domain
- Data-Related Vulnerabilities
- OWASP Top 10 for LLMs 2025
- None mapped
- Classification Reason
- Matched keywords: langgraph
Related Threats
MITRE ATT&CK Enterprise TechniquesAI
Why these techniques?
Vulnerability enables exploitation of remote/internal application services via unsafe msgpack deserialization (CWE-502) to achieve Python code execution after DB write access is obtained.
CVEs Like This One
Affected Assets
Mitigating Controls
Mitigating Controls (NIST 800-53 r5) AI
Requires validation of checkpoint data before msgpack deserialization, directly blocking crafted payloads that trigger unsafe object reconstruction.
Verifies integrity of stored checkpoints and detects unauthorized modifications to the SQLite backing store before they are loaded.
Protects checkpoint data at rest in SQLite, reducing the ability of an attacker to tamper with msgpack payloads even after obtaining privileged DB access.