EUVD-2026-98960
GHSA-crch-2448-h674
Attack ChainAIDerived
Hypothetical attack flow derived from CVE metadata – not a confirmed exploit chain
-
AccessReach DeepSeek V4 endpoint over network
-
ExploitSubmit crafted input request
-
ExecutionBypass access control check
-
ImpactExfiltrate exposed sensitive data
Vulnerability Assessment
| Weakness |
CWE
OWASP Top 10:2025
A01:2025 - Broken Access Control
ASVS 5.0
V14 - Data Protection Outbound data filtering
|
| Exploitation Conditions | Per the supplied CVSS vector (AV:N/AC:L/PR:N/UI:N), exploitation requires only network reachability to a DeepSeek V4 instance/API and a specially crafted input - no authentication, no privileges, and no user interaction are indicated. The description names no required feature flag or non-default configuration, so on the available data this appears to affect reachable deployments generally; the only stated prerequisite is the ability to send a crafted request to the service. Limiting factors cannot be confirmed because affected versions are listed as 'n/a' and no vendor configuration guidance is provided - defenders cannot yet scope which V4 builds or deployment modes are vulnerable. |
| Impact | Signals are mixed and the underlying data is thin. The provided CVSS 3.1 vector (AV:N/AC:L/PR:N/UI:N, C:H/I:H/A:N, base 9.1) describes a remotely reachable, low-complexity, unauthenticated issue against a non-scope-changing component - consistent with the SSVC read (Exploitation: poc, Automatable: yes, Technical Impact: total). Those two sources agree that the flaw is easy to reach and highly impactful. However, EPSS is not provided, so no probabilistic exploitation estimate is available, and the CVE is NOT in CISA KEV, so there is no evidence of active in-the-wild exploitation - only a referenced public gist indicating that exploit/POC material exists. The CWE assignment (CWE-201, information exposure through sent data) and the 'Information Disclosure' tag point to a confidentiality-centric bug, which CONFLICTS with the vector's I:H (high integrity impact): a pure information-disclosure/incorrect-access-control issue would normally show I:N, so the I:H in the input should be verified with the vendor. Net: this reads as a genuinely high-severity, easily-reachable confidentiality flaw worth prompt attention, but 'confirmed critical priority' is overstated given the vague one-line description, placeholder CPE (cpe:2.3:a:n/a:n/a), absent EPSS, and no KEV listing. |
| Exploit ScenarioDerived | A remote, unauthenticated attacker sends a crafted input to a DeepSeek V4 endpoint and, due to incorrect access control, causes the service to return or expose sensitive information it should not disclose. Because the CVSS vector is AV:N/AC:L/PR:N/UI:N and SSVC marks the issue automatable, the request requires no special timing or user interaction and could be scripted at scale. A public gist is referenced, indicating exploit/POC material is available. |
Get an e-mail when this CVE is exploited, patched or re-scored. Free account.
Alert meAnalysisAIv1 – Oct 14, 2026
Crafted input sent over the network to a reachable DeepSeek V4 instance or API can cause the service to return sensitive information to the requester, with the assessed vector requiring no authentication, no privileges, and no user interaction (CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:H/I:N/A:N - confidentiality-only impact). Publicly available exploit code exists: the CVE record links a public GitHub gist alongside the vendor homepage, and CISA's SSVC assessment marks the issue automatable with total technical impact, although the record's 9.1 score implies an integrity impact our own assessment does not confirm. Defenders cannot yet scope exposure precisely because no affected build, version, or deployment mode is published, so the practical risk is highest for any DeepSeek V4 endpoint or API that is reachable from untrusted networks.
RemediationAI
No vendor-released patch identified at time of analysis, so remediation must be compensating. Restrict network reachability to DeepSeek V4 endpoints and APIs to trusted networks or client ranges (side effect: it breaks legitimate remote and third-party integrations), and place an authenticating and authorizing gateway - API gateway or reverse proxy enforcing authentication and per-caller authorization - in front of the service, since the assessed vector is PR:N and no existing access control is implied; this adds latency and operational complexity but directly closes the unauthenticated path. Because the weakness is sensitive information leaving the service in responses (CWE-201), add response inspection or DLP pattern matching over model output and API replies, accepting some false positives and added latency, and log, alert on, and where feasible rate-limit anomalous or repetitive crafted requests (trade-off: throttling can degrade legitimate high-volume use). Treat the linked public gist as available exploit material and prioritize reducing external exposure of any DeepSeek V4 endpoint accordingly. Monitor https://nvd.nist.gov/vuln/detail/CVE-2026-52394 and EUVD-2026-98960 for a scoped fix version, and contact the vendor via http://deepseek.com for patch and configuration guidance.
Affected VersionsENISA
Affects your stack
Find out whether the stacks and servers you run are on an affected version.
Checking this CVE against your own stacks and servers is part of the Basic plan and up.
Technical ContextAI · show details
The root cause class is CWE-201 (Insertion of Sensitive Information Into Sent Data), meaning data that the recipient is not entitled to see is included in output the service sends back, which matches the record's 'incorrect access control' wording and the Information Disclosure tag. This is not a memory-safety or downstream-injection flaw; the crafted input acts as a trigger, and the failure lies in the service's failure to authorize, filter, or redact the sensitive content it places in its response. The CPE supplied with the record is an unpopulated placeholder (cpe:2.3:a:n/a:n/a) and the ENISA EUVD entry (EUVD-2026-98960) lists affected versions as 'n/a n/a', so neither the concrete vendor product identity nor the vulnerable version range is machine-resolvable from this dataset. The reference set consists of the vendor domain (http://deepseek.com), a public gist that appears to carry proof-of-concept material, and the NVD detail page; no dedicated vendor security advisory with version or configuration guidance is present.
Know when the next one hits your stack
- Watch products and get an email when something new is published
- Check this CVE against your own stack and servers
- Exploit and PoC references, threat intelligence from 22 sources
No credit card · 7-day full trial
Severity by source
AV:N/AC:L/PR:N/UI:N/S:U/C:H/I:H/A:N
Crafted-input, network-reachable, unauthenticated flaw (AV:N/AC:L/PR:N/UI:N); CWE-201 is information disclosure, so C:H but I:N/A:N, diverging from the input's I:H.
Primary rating from Vendor (mitre).
CVSS VectorVendor: mitre
Lifecycle Timeline
3DescriptionCVE.org
DeepSeek V4 is vulnerable to Incorrect Access Control via a crafted input.
Executive ViewAI
DeepSeek V4 endpoints and APIs reachable from untrusted networks can disclose sensitive information to unauthenticated requesters, and publicly available exploit code exists, while no vendor-released patch identified at time of analysis. Because no affected build, version, or deployment mode is published, defenders cannot precisely scope exposure, so leadership should treat every externally reachable DeepSeek V4 service as potentially affected. The organizational impact is potential leakage of confidential data, which could harm customer trust, violate privacy or contractual obligations, and invite regulatory scrutiny until exposure is mitigated.
No vendor-released patch identified at time of analysis; interim mitigations include restricting DeepSeek V4 endpoints to trusted networks, placing them behind an authenticated API gateway, applying WAF rules to block exploit patterns, monitoring for sensitive data exfiltration, and disabling the service if it cannot be adequately isolated.
Threat Intelligence Overview
NVD
CVE.org
EPSS
Vulnrichment
How this decision was reached
The engine reads each signal below and applies a fixed list of rules from top to bottom. The first rule that matches sets the recommended action – like a statement, every line is shown.
What every action level means
SSVC AssessmentENISA
Priority Score
Attack Technique
Information disclosure occurs when an application unintentionally exposes sensitive data that aids attackers in reconnaissance or directly compromises security.
Exploit-DB