Anthropic Admin API Key Creation for Persistence

This threat concerns the abuse of administrative credentials within the Anthropic platform. After successfully compromising an administrative account, threat actors may leverage their elevated permissions to generate new Admin API keys. Unlike standard interactive browser sessions protected by Single Sign-On (SSO), these API keys provide durable, programmatic access to sensitive organizational and compliance APIs. Because these keys do not time out, they allow an adversary to maintain presence and automate malicious activities - including exporting sensitive data, modifying roles, or suppressing compliance logs - even after the victim attempts to secure their account by resetting passwords or locking out the primary account from the Identity Provider (IdP). This technique represents a high-impact persistence mechanism that bypasses standard account lifecycle management controls.

Attack Chain

1. Attacker gains unauthorized access to a privileged Anthropic user account (via credential harvesting or session hijacking).
2. Attacker logs into the Anthropic Console using the compromised credentials.
3. Attacker navigates to the API credential management section within the administrative dashboard.
4. Attacker creates a new Admin API key, assigning broad scopes to maximize control-plane access.
5. Attacker stores the newly generated key securely to ensure off-platform persistence.
6. Attacker leverages the API key to perform automated reconnaissance, data exfiltration, or modification of security settings.
7. Attacker uses the programmatic access to maintain activity even if the original compromised interactive account is disabled or revoked by defenders.

Impact

Successful exploitation allows for long-term persistence within an organization's Anthropic instance. Attackers can bypass standard SSO session timeouts, maintain access following account recovery efforts, and automate large-scale data exports or modifications to sensitive compliance logs. This impacts the confidentiality, integrity, and availability of an organization's AI deployment and associated proprietary data.

Recommendation

Prioritize monitoring of administrative API key lifecycle events to identify unauthorized persistence.

• Deploy detection rules to audit all admin_api_key_created events within the Anthropic Audit log stream.
• Establish a formal inventory of approved Admin API keys; cross-reference every new creation event against verified change requests or onboarding tickets.
• Investigate any key creation performed by accounts that recently underwent primary ownership or administrative role changes.
• Upon detecting an unauthorized key creation, immediate response must include revoking the compromised anthropic.audit.admin_api_key_id, rotating all administrative credentials, and auditing API activity associated with the revoked key.