🔥The 9th Round of Easy Loan, Earn $40 Reward is in progress❗️ ⏰ Promotion Period: January 15th - Feburary 15th, 2025 👉 Register now and check more details at gate.io/campaigns/358

Everyone, we have an AI bug bar!

We're hiring!🥳

Nice summer school on security&AI happening at KU Leuven this September

Conformal Prediction is one of the most exciting developments in trustworthy ML. Here's your chance to take part and submit your lovely work to the 13th edition of COPA, which will take place in Milan, from September 9-11. Submission deadline: March 15. copa-conference.com

POTs examples in the wild Bogdan Kulynych Rebekah Overdorf 🏳️‍🌈 Carmela Troncoso (arxiv.org/abs/1806.02711)

Missed the event? Watch it again on our YouTube channel: youtube.com/watch?v=i6h_M2… Stay tuned for future events! Thank you again, Giovanni Cherubin, for talking about your research at our seminar! Hint: find all past events at: pralab.github.io/mlsec/past/

So sad to learn about Ross Anderson 's passing. It's hard to overstate how much he's done for Security Engineering, in such a wide range of problems. His (and his collaborations') works always brought different, no-nonsense, and pragmatic ideas to security. He'll be missed.

Very excited about this work!! LLMs in applications process inputs from many sources, making them vulnerable to prompt injections. We look into models' internals (activations) to catch if models drifted from users' instructions after processing supposedly data-only sources. 1/

We finished 2nd in this LLM CTF competition. Read the report we wrote with the organizers for juicy lessons about why LLM alignment, defensive system prompts, and input and output filters are insufficient to protect the confidentiality of information included in an LLM query.

this👇

Confirmed line-up for KU Leuven PhD School on Security & Privacy & AI: Giovanni Cherubin , @mjuarezm_ , Christian Wressnegger , Fabio Pierazzi , Adriana Sejfia , Dr. Azqa , Peggy Valcke , Sofie Royer. Hosted by Vera Rimmer, Fabio Roli, Lorenzo Cavallaro. cybersecurity-research.be/summer-school-… #AI #security

More info on our LLM ctf soon😉

Come and test your skills at prompt injection! Bring your best attacks

The main phase of the competition has ended today. We have received over 370K submissions!!! 🥳🤯🫨 We are grateful for and blown away by the active participation and kind feedback we have received :) Please keep checking the website for follow-ups: llmailinject.azurewebsites.net

More to come soon. Watch out this space! llmailinject.azurewebsites.net

#kolmogorov

Looks like folks had fun with LLMail-Inject, so here's a new updated version! Launching next week