Severity by source
CVSS:4.0/AV:N/AC:L/AT:N/PR:N/UI:N/VC:H/VI:H/VA:N/SC:N/SI:N/SA:N/E:X/CR:X/IR:X/AR:X/MAV:X/MAC:X/MAT:X/MPR:X/MUI:X/MVC:X/MVI:X/MVA:X/MSC:X/MSI:X/MSA:X/S:X/AU:X/R:X/V:X/RE:X/U:X
Unauthenticated network access via self-registration justifies PR:N/AV:N/AC:L; overwriting a shared singolo config to hijack all users' data is a scope change (S:C) with high C and I but no availability impact.
Primary rating from Vendor (VulnCheck).
CVSS VectorVendor: VulnCheck
Lifecycle Timeline
1DescriptionCVE.org
Cognee before 1.2.0 contains an improper access control vulnerability that allows unauthenticated attackers to overwrite the global LLM provider configuration by self-registering an account and calling the settings endpoint, which performs no admin or superuser check. Attackers can redirect all LLM operations instance-wide to an attacker-controlled endpoint by exploiting the process-wide singleton configuration cache, enabling exfiltration of prompts, uploaded documents, extracted entities, and knowledge graph content from all users.
AnalysisAI
Instance-wide LLM configuration hijacking in Cognee before 1.2.0 lets any remote attacker self-register a normal account and overwrite the global LLM provider settings through the /api/v1/settings endpoint, which performs no admin or superuser authorization check. Because the configuration is held in a process-wide singleton cache, a single call redirects every user's LLM operations to an attacker-controlled endpoint, exfiltrating prompts, uploaded documents, extracted entities, and knowledge-graph content. This is confirmed publicly available exploit code exists (reported by VulnCheck, with a released vendor patch in v1.2.0), and the CVSS 4.0 score is 9.3 (Critical).
Technical ContextAI
Cognee is an open-source AI memory/knowledge-graph framework (vendor topoteretes) that ingests documents and builds knowledge graphs by delegating operations to an external LLM provider. The affected component is the settings management API (per the VulnCheck advisory, api/v1/settings), which mutates the provider/base-URL/model/key configuration used by all LLM calls. The root cause is CWE-862 (Missing Authorization): the endpoint updates a sensitive, global setting without verifying the caller is an admin or superuser. The impact is amplified by an architectural choice - the LLM configuration is stored in a process-wide singleton cache shared across all sessions and users - so one write by a low-privilege self-registered account changes behavior for the entire instance rather than just the attacker's own scope.
RemediationAI
Vendor-released patch: upgrade Cognee to v1.2.0 or later (https://github.com/topoteretes/cognee/releases/tag/v1.2.0); the fix is in commit d10b1b77e2157c6238fd4d1acb1923a048991699, which adds the missing admin/superuser authorization check to the settings endpoint. If you cannot upgrade immediately, disable open/self-registration so unauthenticated users cannot obtain an account to reach the endpoint (trade-off: breaks any legitimate self-service onboarding), and restrict network access to the /api/v1/settings endpoint via a reverse proxy or WAF rule allowing only trusted admin IPs (trade-off: requires that admins still be able to reach it for legitimate configuration). Additionally, pin and monitor the configured LLM provider base URL/endpoint and alert on unexpected changes so a hijack is detected quickly. Review the advisory at https://www.vulncheck.com/advisories/cognee-unauthorized-llm-configuration-overwrite-via-api-v1-settings for details.
Same weakness CWE-862 – Missing Authorization
View allSame technique Authentication Bypass
View allShare
External POC / Exploit Code
Leaving vuln.today
EUVD-2026-42095
GHSA-49f7-whx5-4256