Mistral AI Source Code Stolen in Data Breach
A hacker group called TeamPCP claimed to have stolen 450 repositories and approximately 5 GB of source code from French AI company Mistral AI, reportedly demanding $25,000 or threatening to leak the data for free. Mistral confirmed its development kits were compromised but stated core systems and user data were unaffected. The company attributed the incident to a TanStack supply chain attack affecting developer environments.
What happened
Hacker group TeamPCP claimed to have stolen 450 repositories and approximately 5 GB of source code from Mistral AI, including code for training, fine-tuning, benchmarking, model delivery, and experiments.
TeamPCP demanded $25,000 for the stolen data and threatened to release it for free on the dark web if no buyer emerged within a week.
Mistral AI confirmed that its development kits were contaminated and that an automated worm associated with a TanStack supply chain attack compromised NPM and PyPi packages published between May 11 and 12, 2026, targeting developer environments.
Reported impact
Mistral AI stated that its core systems and user data were unaffected by the incident, and that there was no evidence its internal infrastructure was breached.
Index analysis
The compromised packages deployed malware on Linux systems and harvested credentials from common locations in developer environments.
Governance controls
Mistral AI advised users to stop using affected package versions immediately and to clean systems where one of these packages had been installed.
Classification
Sources and evidence
This record was researched and written by the Index, based on the following original reporting.