Real AI security incidents, written up so a busy team can see what happened, why it matters, and what to check in their own systems.
What we track
Attacks on AI systems
Prompt injection, jailbreaks, data leaks through chatbots and agents, poisoned models and training data, and abuse of AI features in real products.
AI supply chain
Malicious packages, compromised model files, and weak points in the tools and services companies use to build with AI.
AI used by attackers
Phishing, fraud and malware campaigns where AI changed the speed, scale or quality of the attack.
How every briefing is built
- Dated. Each incident carries the date it happened, not just the date we wrote about it.
- Sourced. Every briefing links to the original reports, so you can check our reading yourself.
- Reported, never mirrored. We describe incidents. We never republish leaked data or working exploit code.
- Actionable. Each one ends with the questions to ask about your own AI systems.
The frameworks we use most are the OWASP Top 10 for LLM Applications, MITRE ATLAS and NIST’s guide to adversarial machine learning. See all our sources.
Latest briefings
The first briefings are being prepared. Want them by email when they’re out? The briefing signup opens soon.