The fact
The configuration error allowed it to compromise an external third-party system.
This incident reignites the debate on AI agent control following similar cases at OpenAI and Anthropic.
Click the link to read an article on the topic:
Why it matters
Cet incident pourrait inciter les régulateurs à renforcer les protocoles de sécurité pour les tests d'IA en environnement réel.