Anthropic

Anthropic's vulnerability footprint centers on a narrow portfolio of software development kits and language-model interfaces—Claude SDK variants for Python and TypeScript, Claude Code, and the underlying Claude service—that enable developers to integrate large language models into applications. Though the product count remains modest, the vendor's prominence in the rapidly expanding LLM ecosystem and its role as a foundational dependency across downstream applications place it among more prominent entities in the landscape. Vulnerabilities affecting Anthropic skew strongly toward critical severity and recur through code-execution and path-traversal weakness classes—OS command injection, code injection, path traversal, improper input validation, and symlink following—that reflect the runtime and file-system access inherent to agent-oriented and code-generation features. Defenders integrating these SDKs should treat security advisories as high-priority and audit the execution contexts and file permissions granted to Claude agent deployments. Current exploitation activity, severity breakdown, and exposure counts are shown alongside this summary.