good lord. at the baggage belts, even #acl2025 #acl2025nlp

First keynote at LLMSEC 2025, ACL: "A Bunch of Garbage and Hoping: LLMs, Agentic Security, and Where We Go From Here" Erick Galinkin Friday 09.05 Hall B Details: sig.llmsecurity.net/workshop/ - #ACL2025NLP

Come to LLMSEC at ACL & hear Niloofar's keynote "What does it mean for agentic AI to preserve privacy?" - Niloofar (✈️ ACL), Meta/CMU (Friday 1st Aug, 11.00; Austria Center Vienna Hall B) See you there! #acl2025 #acl2025nlp

We have put up all slide decks on the tutorial website: llm-guardrails-security.github.io 🥳🥳🥳 Although I was only able to deliver the tutorial remotely due to visa constraints, I was really thrilled to learn that our tutorial received a quite full room of audience for the entire 3.5 hr!

At ACL in Vienna? Hear the world's leading prompt injector talk at LLMSEC on Friday! Johann Rehberger Johann Rehberger will be presenting the afternoon keynote at 14.00 in Hall B > sig.llmsecurity.net/workshop/ #ACL2025NLP #ACL2025

From the AI workshop I'm in: "The S in MCP stands for security"

on ed hovy's framing of LLM heavy papers at #ACL2025NLP being "popcorn papers", which might seem fine enough while you're consuming them, but are not filling or nutritious, and just leave one hungry for some real knowledge 1. it's correct

LLMSEC proceedings are up! sig.llmsecurity.net/proceedings.pdf (Anthology is processing) #ACL2025NLP

Amazing LLMSEC keynote by Johann Rehberger. Yet another call to focus on people - "we need humans to lead AI, not AI to lead humans" #acl2025nlp #acl2025

aaand done. that was a lovely acl. time to go home. wedding tomorrow, moving house sunday, back to work monday, sleep pushed til q4 😄 #acl2025

Brutally honest aesthetics opinion, apologies: One pic feels like hardened stainless steel running COBOL, reads Hartford, Connecticut U.S.A. on a name plate somewhere. The other feels like recycled ABS plastic javascript shipped out of Temu. We have all this tech, but no aura.

you may live to see man-made horrors beyond your comprehension

he made my fortune

really satisfied that "Importing Phantoms: Measuring LLM Package Hallucination Vulnerabilities" got accepted to CAMLIS. package hallucination is so, so real and the bigger models are worse at it. ms here: arxiv.org/abs/2501.19012 work led by Arjun Krishna , with Erick Galinkin

moved house & unpacking, found this uncommon relic. feeling pretty prepared this time tho

he's not wrong 😍

little known fact: in the time of word2vec there were so few weights, no huggingface, we frequently typed them by hand.