Adam Gleave - AI2050

Adam Gleave - AI2050 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

 
 

 
 
 
 
 
 
 
 
 
 

 

 

 

 
 
 

 

 
 
 
 
 Fellows Community 
 
 
 
 
 
 
 
 
 
 

 
 
 
 Back 
 
 
 

 
 

 
 Affiliation 
 Co-founder & CEO, FAR.AI

 Hard Problem 
 Assurance

 
 

 
 Adam Gleave

 
 
 2025 Early Career Fellow 

 
 Adam Gleave is the co-founder and CEO of FAR.AI, an AI safety research institute working to ensure advanced AI is safe and beneficial for everyone. Adam’s research focuses on securing advanced AI systems. Outside of FAR.AI, Adam is a board member of the Safe AI Forum (SAIF), Model Evaluation and Threat Research (METR), and the London Initiative for Safe AI (LISA), and an advisor for Timaeus and the AI Risk Mitigation Fund. Prior to founding FAR.AI, Adam received his PhD from UC Berkeley under the supervision of Stuart Russell, and previously worked at Google DeepMind with Jan Leike and Geoffrey Irving and several quantitative trading firms. 

 AI2050 Project 

 Gleave’s project develops techniques to detect and eliminate hidden behaviors in advanced AI systems. Just as security researchers find and fix vulnerabilities in software, they’re creating methods to audit AI models for concealed objectives that could lead to harmful actions. Through a “red-team/blue-team” approach, they’ll first create models with sophisticated hidden behaviors, then develop tools to identify and remove them. This work addresses risks from both malicious actors inserting backdoors and unintentional AI misalignment. The resulting methods will help ensure that increasingly powerful AI systems remain transparent and trustworthy, allowing society to benefit from AI advances while managing potential risks. 

 
 Affiliation 
 Co-founder & CEO, FAR.AI

 Hard Problem 
 Assurance