Security Now (Audio) · Booby Trapping LLMs - AI Frenzy in Cybercrime
Audio streams from the publisher’s own feed. Verbatim does not host or copy it.
embedded a URL pointing to a remote payload inside a poisoned model so the model can fetch and execute malicious code on trigger
“And if you use a logistic regression on top of embeddings or a small model, you get the intelligence of logistic regression on top of embeddings or a small model.”
“Torquay and all the other models learn to generalize maliciously, figure two.”
“If I were 18, 19 now, I would say, what's a foundation model I could build for a field I'm interested in? ... I'd like to build a foundation model for architecture ... I can query it to build me a house, analyze this, where are the structural points, and so forth.”
“There is an agent role called the cartographer... that creates a threat model. You can do this in two ways... threat model after the code is created”
The strongest new moments on your topics, in the speakers' own words. Free for three topics.
No card. No password. One-click unsubscribe.
Pro watches up to 25 names or phrases and emails you the morning after they are said.
See Pro, $29 a monthVerbatim quotes a short extract and links to the episode; the episode belongs to its publisher. Remove a show.