Raw vector
CVSS:4.0/AV:N/AC:L/AT:N/PR:N/UI:P/VC:L/VI:L/VA:L/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:XSummary
CVE-2026-21852 is a medium-severity Insufficiently Protected Credentials (CWE-522) vulnerability in Anthropic Claude Code. Its CVSS base score is 5.3 (Medium).
Operationally, exploitation aligns with the MITRE ATT&CK technique Unsecured Credentials (T1552); ranked in the top 2% of CVEs by exploit likelihood; it is not currently listed in the CISA KEV catalog.
This vulnerability is AI-related — categorised as Enterprise AI Assistants; in the Supply Chain and Deployment risk domain.
The strongest mitigations our analysis identified map to SC-28 (Protection of Information at Rest) and SC-8 (Transmission Confidentiality and Integrity) — see the control section below for these in your framework.
Deeper analysis AI-assisted summary
Synthesised by an AI model from the NVD description and linked references — a reading aid, not an authoritative source.
CVE-2026-21852 is a vulnerability in Claude Code, an agentic coding tool developed by Anthropic, affecting versions prior to 2.0.65. The issue resides in the project's load flow, where malicious repositories can exfiltrate sensitive data, including users' Anthropic API keys, before a trust confirmation prompt is displayed. Specifically, an attacker-controlled repository can include a settings file that overrides the ANTHROPIC_BASE_URL environment variable to point to an attacker-controlled endpoint. Upon opening the repository, Claude Code reads this configuration and issues API requests immediately, bypassing user approval. The vulnerability is rated with a CVSS v3.1 base score of 7.5 (AV:N/AC:L/PR:N/UI:N/S:U/C:H/I:N/A:N) and is associated with CWE-522 (Insufficiently Protected Credentials).
Attackers can exploit this vulnerability remotely with no privileges or user interaction beyond convincing a target to open a malicious repository, such as via social engineering through shared links or recommendations. No authentication is required, making it accessible to any remote adversary. Successful exploitation allows the attacker to capture API requests containing the victim's Anthropic API keys, enabling unauthorized access to the victim's Anthropic services, potential further compromise of associated accounts, and exfiltration of additional data processed through those APIs.
The security advisory at https://github.com/anthropics/claude-code/security/advisories/GHSA-jh7p-qr78-84p7 recommends updating to version 2.0.65 or later, which includes a patch addressing the configuration override issue. Users on the standard auto-update channel have already received the fix, while those using manual updates must apply it explicitly to mitigate the risk.
OWASP Top 10 for Web (2025)
EU & UK References
- 🇪🇺 ENISA EUVD: EUVD-2026-3597
Vulnerability Data
Claude Code is an agentic coding tool. Prior to version 2.0.65, vulnerability in Claude Code's project-load flow allowed malicious repositories to exfiltrate data including Anthropic API keys before users confirmed trust. An attacker-controlled repository could include a settings file that…
more
sets ANTHROPIC_BASE_URL to an attacker-controlled endpoint and when the repository was opened, Claude Code would read the configuration and immediately issue API requests before showing the trust prompt, potentially leaking the user's API keys. Users on standard Claude Code auto-update have received this fix already. Users performing manual updates are advised to update to version 2.0.65, which contains a patch, or to the latest version.
- CWE(s)
AI Security AnalysisAI
- AI Category
- Enterprise AI Assistants
- Risk Domain
- Supply Chain and Deployment
- OWASP Top 10 for LLMs 2025
- None mapped
- Classification Reason
- Matched keywords: anthropic, claude
Related Threats
MITRE ATT&CK Enterprise Techniques
CVEs Like This One
Affected Assets
Mitigating Controls
Control response
—
—
- 1 hardening rule · 1 OS baseline
V11.3.3
Mitigating Controls (NIST 800-53 r5) AI
Protection of information at rest requires encryption or equivalent safeguards for stored credentials.
Transmission confidentiality and integrity enforcement stops credentials from being sent in plaintext or without protection.
Authenticator management requires secure generation, storage, and distribution of credentials, directly stopping insecure methods.
Mitigating Controls (NIST CSF 2.0) AI
Derived directly from the weakness types (CWEs) cited in the NVD entry via our AI-authored CWE→CSF cross-walk (authority under review) — links open the control.
Encrypting data-at-rest fully prevents insecure credential storage while only partially satisfying the broader data-protection outcome.
Encrypting data-in-transit fully prevents interception of credentials in motion while only partially satisfying the broader data-protection outcome.
Credential management practices directly reduce insecure storage/transmission but do not guarantee encryption or transport protection.
Protecting identity assertions covers conveyance of credentials but is narrower than full credential lifecycle protection.
Authentication policies can enforce stronger credential handling yet address only verification, not storage or transit protection.
Mitigating Controls (ISO/IEC 27001:2022 Annex A) AI
Derived directly from the weakness types (CWEs) cited in the NVD entry via our AI-authored CWE→ISO cross-walk (authority under review) — links open the control.
Requiring protected storage, transmission, and non-display of passwords prevents credentials from being stored or sent in clear text where they can be harvested by unauthorized actors.
Protecting secret and private keys against disclosure and unauthorized use decreases the exposure of credentials that are stored or transmitted in recoverable form.
Forbidding clear-text transmission and display of passwords, plus the use of stronger alternatives to passwords, prevents credentials from being obtained or reused by attackers.
Acceptable-use expectations that cover protection of credentials and information assets throughout their lifecycle discourage practices that expose or mishandle authentication material.
Contractual clauses that survive termination help ensure that credentials and other secrets are not retained or misused after employment ends.
Regular reminders about password security and personal accountability make users less likely to store or transmit credentials in cleartext or other unprotected forms.