Open-WebUI versions up to 0.8.10 contain an Insecure Direct Object Reference (IDOR) flaw in the notes API — any authenticated user can read another user's private notes simply by cycling through UUID-formatted note IDs with no authorization check on the server side. In enterprise or team deployments of Open-WebUI (a popular frontend for local LLM serving), notes frequently contain sensitive system prompts, proprietary configurations, internal business data, or credentials, making the real business impact substantially higher than the CVSS 6.5 score implies. A secondary design flaw compounds the risk: even when administrators disable the Notes feature via settings, a logged-in attacker can re-enable it client-side by intercepting the /api/config response, meaning the disabling control provides no meaningful defense. Upgrade immediately to open-webui v0.8.11, which enforces server-side ownership checks on all note endpoints; there is no safe workaround on earlier versions.
What is the risk?
CVSS 6.5 (Medium) understates business risk in multi-user enterprise deployments. The attack requires only a valid account (Privileges Required: Low), has no complexity barrier (Attack Complexity: Low), and needs no user interaction — trivially scriptable. Confidentiality Impact is rated HIGH in the CVSS vector, correctly reflecting that all notes across the instance are accessible. The administrative disable-bypass via /api/config manipulation eliminates a natural compensating control. With 91 prior CVEs in open-webui, this package has a documented pattern of security weaknesses. No public exploit code or Nuclei template exists yet, and no CISA KEV listing, but the low skill floor means exploitation is realistic shortly after this advisory becomes widely known.
Attack Kill Chain
What systems are affected?
| Package | Ecosystem | Vulnerable Range | Patched |
|---|---|---|---|
| open-webui | pip | <= 0.8.10 | 0.8.11 |
Do you use open-webui? You're affected.
Severity & Risk
Attack Surface
What should I do?
5 steps-
PATCH
Upgrade open-webui to v0.8.11 or later immediately — this is the only complete fix.
-
DETECT
Review access logs for high-frequency or sequential requests to GET /api/v1/notes/{id} from a single authenticated session; UUID enumeration typically shows patterned or random ID variation with a single source token.
-
INTERIM CONTROL
If patching is delayed, restrict network-level access to the Open-WebUI API to trusted internal segments to limit the attacker pool to internal accounts only.
-
AUDIT
Examine existing notes for sensitive content (credentials, PII, proprietary configs) that may have been exposed; treat any multi-user instance on <= 0.8.10 as potentially compromised for the notes feature.
-
NOTIFY
If your instance is regulated (healthcare, finance, public sector), assess whether a data breach notification obligation is triggered by the potential cross-user note exposure.
Classification
Compliance Impact
This CVE is relevant to:
Frequently Asked Questions
What is CVE-2026-45666?
Open-WebUI versions up to 0.8.10 contain an Insecure Direct Object Reference (IDOR) flaw in the notes API — any authenticated user can read another user's private notes simply by cycling through UUID-formatted note IDs with no authorization check on the server side. In enterprise or team deployments of Open-WebUI (a popular frontend for local LLM serving), notes frequently contain sensitive system prompts, proprietary configurations, internal business data, or credentials, making the real business impact substantially higher than the CVSS 6.5 score implies. A secondary design flaw compounds the risk: even when administrators disable the Notes feature via settings, a logged-in attacker can re-enable it client-side by intercepting the /api/config response, meaning the disabling control provides no meaningful defense. Upgrade immediately to open-webui v0.8.11, which enforces server-side ownership checks on all note endpoints; there is no safe workaround on earlier versions.
Is CVE-2026-45666 actively exploited?
No confirmed active exploitation of CVE-2026-45666 has been reported, but organizations should still patch proactively.
How to fix CVE-2026-45666?
1. PATCH: Upgrade open-webui to v0.8.11 or later immediately — this is the only complete fix. 2. DETECT: Review access logs for high-frequency or sequential requests to GET /api/v1/notes/{id} from a single authenticated session; UUID enumeration typically shows patterned or random ID variation with a single source token. 3. INTERIM CONTROL: If patching is delayed, restrict network-level access to the Open-WebUI API to trusted internal segments to limit the attacker pool to internal accounts only. 4. AUDIT: Examine existing notes for sensitive content (credentials, PII, proprietary configs) that may have been exposed; treat any multi-user instance on <= 0.8.10 as potentially compromised for the notes feature. 5. NOTIFY: If your instance is regulated (healthcare, finance, public sector), assess whether a data breach notification obligation is triggered by the potential cross-user note exposure.
What systems are affected by CVE-2026-45666?
This vulnerability affects the following AI/ML architecture patterns: LLM chat interfaces, self-hosted AI productivity platforms, enterprise local LLM deployments, multi-user AI sandboxes.
What is the CVSS score for CVE-2026-45666?
CVE-2026-45666 has a CVSS v3.1 base score of 6.5 (MEDIUM).
Technical Details
NVD Description
### Summary The API /api/v1/notes/{note_id} endpoint lacks proper authorization checks, allowing authenticated users to retrieve notes belonging to other users by guessing or enumerating UUIDs. This results in unauthorized disclosure of potentially sensitive or private user data. ### Details - if notes is enabled from UI (Settings >> General >> Features >> Notes (Beta)) - From API, attacker can access other user notes - if notes is disabled from UI (Settings >> General >> Features >> Notes (Beta)) - Then attacker can enable the notes from /api/config and access other user notes ### PoC - Step 1: Log in to the application as a valid user (User A).  - Step 2: Intercept or inspect the response from the endpoint GET /api/config.  - Step 3: Observe the field "enable_notes": false in the JSON response. - Step 4: Manually change "enable_notes" to true using browser DevTools or by intercepting and modifying the response via a proxy like Burp Suite. (Please note, the occurrence of this API comes twice, hence modification needs to be made twice as well.)  - Step 5: Observe the loaded frontend application; the previously hidden notes form will now be visible.   - Step 6: Again, click on any note, intercept the request, and Replace the note_id in the URL with a different note ID known to belong to another user (e.g., by guessing or bruteforcing). - Step 7: Send the modified request while remaining logged in as User A. - Step 8: Observe that the server returns the content of another user's note, confirming unauthorized access.   ### Impact 1. Unauthorized access to user-created notes 2. Possible exposure of confidential or sensitive uploaded data 3. Violation of user privacy and data isolation 4. High risk of legal or compliance breaches in regulated environments ## Resolution Fixed in commit [de3317e26](https://github.com/open-webui/open-webui/commit/de3317e26bb67a2a7ea015a183bbd1d369880ebd), first released in **v0.8.11** (Mar 2026). All per-id note endpoints (`GET /api/v1/notes/{id}`, `POST /api/v1/notes/{id}/update`, `POST /api/v1/notes/{id}/access/update`, deletion) now enforce ownership: the handler fetches the note, then requires the caller to be admin, the note owner, or have an `AccessGrants` grant for the appropriate permission (`read` for retrieval, `write` for mutation). A non-owner with no grant receives 403. Users on `>= 0.8.11` are not affected.
Exploitation Scenario
An attacker with a low-privilege Open-WebUI account — for example, a contractor, intern, or external researcher with trial access — first checks /api/config via the browser developer tools and observes 'enable_notes': false. They intercept the config response with Burp Suite, flip the flag to true, and the Notes UI becomes visible in their session. They then click any note link in their own session to capture a sample UUID-formatted note_id from the GET /api/v1/notes/{id} request. From there, a simple script iterates through UUID variants (or uses knowledge of a target user's ID range) sending authenticated GET requests; each request that returns 200 instead of 403 reveals another user's private note. Across a 500-user enterprise instance, an attacker could exfiltrate the entire notes corpus within minutes, harvesting system prompt designs, pasted API keys, or confidential business data with no server-side resistance.
Weaknesses (CWE)
CVSS Vector
CVSS:3.1/AV:N/AC:L/PR:L/UI:N/S:U/C:H/I:N/A:N References
Timeline
Related Vulnerabilities
CVE-2026-44551 9.1 open-webui: LDAP auth bypass — full account takeover
Same package: open-webui CVE-2026-45672 8.8 open-webui: code exec gate bypass via API endpoint
Same package: open-webui CVE-2026-44552 8.7 open-webui: Redis cache poisoning enables cross-instance tool hijack
Same package: open-webui CVE-2025-64495 8.7 Open WebUI: XSS-to-RCE via malicious prompt injection
Same package: open-webui CVE-2026-45315 8.7 Analysis pending
Same package: open-webui