Writing
ai security
3 articles on this subject.
Indirect prompt injection: EchoLeak, CVE-2025-32711, and the email nobody opened
A crafted email sitting unread in an inbox was enough to make Microsoft 365 Copilot exfiltrate internal data. The injectable surface is every document the model reads, and the control that bounds it is least privilege on tools.
7 min read
Pickle is a code format, not a data format: CVE-2025-32434 and the model file you just downloaded
torch.load(weights_only=True) was the recommended safe path, and it was bypassable. A practical look at Python deserialization, pickle, PyYAML and jsonpickle, how one gadget works, and what actually stops it.
12 min read
An LLM is not a security boundary: building a moderation layer, then walking through it
We built a kids' chat app where a language model decides which messages are safe, then attacked the decision. Prompt injection is OWASP's number one LLM risk for the second edition running, and the reason is structural.
8 min read