Overview
Severity: CRITICAL | Affected: Cognition Corp | Category: breach
Cognition Corp, a leading developer of autonomous AI agents, disclosed a significant security breach on June 25. Attackers exploited a zero-day vulnerability in a popular MLOps platform used internally, exfiltrating gigabytes of sensitive data. The stolen assets include the proprietary weights of their flagship 'Cog-7' model and a substantial portion of its curated training dataset, which reportedly contains sensitive corporate information from enterprise clients used for fine-tuning. The breach poses a severe risk, as malicious actors could replicate the model, develop countermeasures, or launch more sophisticated attacks. The company has notified affected customers and is working with cybersecurity firms to investigate the incident. The leak of both model weights and training data represents a critical blow to the company's intellectual property and the security of its clients' information, highlighting the growing importance of securing the entire AI development lifecycle.