Mistral denies a fresh security breach, but the code on sale looks a lot like May's leak (www.techradar.com)

Mistral AI has faced accusations of a new security breach after a seller on a cybercrime forum claimed to be offering the company's full source code. While the seller, known as "mrwho," alleges this is the result of a breach occurring after a significant attack in May 2026 by the TeamPCP hacking group, Mistral's internal investigation has found no evidence to support such claims. The company remains cautious, stating that no customer data has been compromised, and crucially, none of the analyzed code appears to contain updates or files created after the May incident, suggesting the supposed leak may be a rehash of earlier compromised data. This incident is particularly noteworthy for the AI/ML community, highlighting ongoing security vulnerabilities in tech companies and their software supply chains. Mistral AI previously suffered significant damage from the May breach, where compromised SDK packages spread malware, leading to systemic exploits. If the September claim proves to be legitimate and contains fresh data, it would signify deeper issues in Mistral's security protocols. However, if the contents mirror the earlier leak, it underscores the challenge of managing and safeguarding proprietary code from malicious actors, raising questions about industry standards in software security amidst rising cyber threats.