AI Supply Chain Vulnerability Exposed

LiteLLM breach reveals AI middleware as critical vulnerability. Attackers bypass models to access data through compromised API gateways. Salt Security Platform offers detection solutions.

Mercor AI Data Breach Exposes Secrets

Mercor, a $10 billion startup providing AI training data for Meta, OpenAI, Anthropic, and Google, suffered a breach after TeamPCP compromised Python library LiteLLM. Four terabytes of data were allegedly stolen, including proprietary training methodologies.

Meta Pauses Mercor Partnership After Breach

Meta and other major AI labs are investigating a security breach at Mercor, a data vendor, that may have exposed sensitive information about AI model training processes.