Mercor Data Breach Tied to LiteLLM Supply Chain Attack Raises Concerns for Anthropic and AI Partners

AI recruiting platform Mercor has confirmed a data breach that security researchers linked to an attack on LiteLLM, an open source library widely used by AI application developers to interface with large language model APIs, raising concerns about the exposure of sensitive data across Anthropic and its partner ecosystem, Moneycontrol reported. LiteLLM is a popular abstraction layer that lets developers call models from Anthropic, OpenAI, Google, and other providers through a unified interface. Its broad adoption in AI application stacks means that a compromise of the library can potentially propagate across many downstream applications and the organizations that use them. Mercor, which uses AI to match candidates to technical roles and counts several AI labs and technology companies among its clients, said the breach affected customer data. The company confirmed the incident but did not disclose the full scope of what was accessed. The attack vector via a trusted open source developer tool follows the pattern of other high-impact supply chain incidents. Developers who have integrated LiteLLM into production systems are advised to audit their dependencies, review recent version changelogs for signs of unauthorized modification, and rotate any API keys or credentials that may have been accessible in affected environments. The incident adds to growing concern about the security posture of the AI application ecosystem, where rapid development and heavy reliance on open source tooling has sometimes outpaced security review practices. LiteLLM maintainers have not yet issued a public statement at time of reporting.