www.scientificamerican.com/article/hack... New article by Deni Bechard at Scientific America covering our work on hijacking Multimodal computer agents published on Arxiv earlier this year. A massive effort by Lukas Aichberger, supported by myself Yarin Gal, Philip Torr, FREng, FRS & Adel Bibi

Excited to share our paper: "Chain-of-Thought Is Not Explainability"! We unpack a critical misconception in AI: models explaining their steps (CoT) aren't necessarily revealing their true reasoning. Spoiler: the transparency can be an illusion. (1/9) 🧵

Youtuber Sabine Hossenfelder has picked up on our paper on how to fool AI agents, you can watch her desrcibing our work here: www.youtube.com/watch?v=KY7_....

🚨 New paper alert: Our recent work on LLM safety has been accepted to ICLR 2025 🇸🇬 We propose a new framework for LLMs safety. 🧵 (1/7) #LLM #AISafety #ICLR2025 #Certification #AdversarialRobustness #NLP #Shhhhhh #DomainCertification #AI

Excited to be working with the UK Excited to be work with the UK AI Security Institute, even more important than normal in these turbulent times. www.linkedin.com/posts/philip...

⚠️ Beware: Your AI assistant could be hijacked just by encountering a malicious image online! Our latest research exposes critical security risks in AI assistants. An attacker can hijack them by simply posting an image on social media and waiting for it to be captured. [1/6] 🧵

🚨 New Paper Alert: Open Problem in Machine Unlearning for AI Safety 🚨 Can AI truly "forget"? While unlearning promises data removal, controlling emergent capabilities is a inherent challenge. Here's why it matters: 👇 Paper: arxiv.org/pdf/2501.04952 1/8

🧵 [1/3] Heading to #Vancouver 🇨🇦 tomorrow to present our latest work in @OxfordTVG #UniversityOfOxford at #NeurIPS2024 🧠: - 💥 Improving on #StylizedImageNet, use #IllusionBench: can you see the cat 🐈‍⬛ Hidden in Plain Sight in the picture 🖼️? Paper: arxiv.org/abs/2411.06287