News – FAR.AI

News – FAR.AI 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

 
 

 
 

 
 
 
 
 
 
 
 
 
 
 
 
 We updated our website and would love your feedback! 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 Events 
 
 
 
 
 
 Events 
 
 
 
 
 
 
 
 
 
 
 
 
 
 Programs 
 
 
 
 
 
 Programs 
 
 
 
 
 
 
 
 
 Blog 
 
 
 
 
 
 About 
 
 
 
 
 
 About 
 
 
 
 
 
 
 
 
 
 Careers Donate 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 Updates on our research, events, and more!

 
 ‍

 
 Sign up for our newsletter 
 
 
 
 
 
 
 

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 Topic 
 
 
 
 
 
 
 
 Event 
 Robustness 
 Interpretability 
 Model Evaluation 
 Alignment 
 
 
 
 
 
 
 
 Authors 
 
 
 
 
 
 
 
 Jean-François Godbout 
 Lars Yencken 
 Matthew Kowal 
 Chris Cundy 
 Euan McLean 
 Dillon Bowen 
 Ann-Kathrin Dombrowski 
 Tony Wang 
 Niki Howe 
 Kellin Pelrine 
 Ethan Perez 
 Claudia Shi 
 ChengCheng Tan 
 Tom Tseng 
 Mohammad Taufeeque 
 Ian McKenzie 
 Adrià Garriga-Alonso 
 Hannah Betts 
 Adam Gleave 
 
 
 
 
 
 
 
 
 
 Thank you! Your submission has been received! 
 
 
 Oops! Something went wrong while submitting the form. 
 
 
 
 
 
 
 
 _ 
 results 
 
 
 
 
 
 
 
 
 
 
 
 Revisiting Frontier LLMs’ Attempts to Persuade on Extreme Topics: GPT and Claude Improved, Gemini Worsened

 
 
 Model Evaluation

 
 
 
 
 We test recently released models from frontier companies to see whether progress has been made on their willingness to persuade on harmful topics like radicalization and child sexual abuse. We find that OpenAI’s GPT and Anthropic’s Claude models are trending in the right direction, with near zero compliance on extreme topics. But Google’s Gemini 3 Pro complies with almost any persuasion request in our evaluation, without jailbreaking.

 
 
 February 11, 2026

 
   

 
 
 
 
 
 
 
 
 
 Matthew Kowal 
 
 
 Jasper Timm 
 
 
 Jean-François Godbout 
 
 
 Thomas Costello 
 
 
 Siao Si Looi 
 
 
 
 
 
 
 
 
 
 
 FAR.AI Selected to Lead EU AI Act CBRN Risk Consortium

 
 
 

 
 
 
 
 FAR.AI has been selected by the European Commission's AI Office to conduct technical safety research supporting the implementation of the EU's landmark Artificial Intelligence Act. We'll tackle one of the most critical safety challenges posed by advanced AI systems: preventing misuse of AI systems to help produce Chemical, Biological, Radiological, and Nuclear (CBRN) threats.

 
 
 February 3, 2026

 
   

 
 We're pleased to share that FAR.AI has been selected by the European Commission's AI Office to conduct technical safety research supporting the implementation of the EU's landmark Artificial Intelligence Act. We'll tackle one of the most critical safety challenges posed by advanced AI systems: preventing misuse of AI systems to help produce Chemical, Biological, Radiological, and Nuclear (CBRN) threats. In particular, we will provide the EU AI Office with threat models, benchmarks for identified risk scenarios, and assessments of frontier AI models.

 We will le

... (truncated, 98 KB total)