Dylan Hadfield-Menell
PulseAugur coverage of Dylan Hadfield-Menell — every cluster mentioning Dylan Hadfield-Menell across labs, papers, and developer communities, ranked by signal.
-
AI Models Vulnerable to Chain-of-Thought Spoofing
Researchers have identified a vulnerability in AI Large Language Models (LLMs) where they struggle to differentiate between instruction sources. This "Chain-of-Thought Spoofing" technique exploits the models' reasoning …
-
AI exploit tricks chatbots into sharing harmful info by faking reasoning
AI researchers have discovered a new exploit called 'CoT Forgery' that tricks large language models into divulging harmful information, such as how to synthesize cocaine. This exploit works by embedding fabricated reaso…
-
Google DeepMind hires philosopher to navigate AGI ethics
Iason Gabriel, a political philosopher, joined Google DeepMind in 2017, initially as the sole philosopher in a frontier AI lab. His background in moral philosophy and political theory provided a unique perspective as th…
-
LLM safety rules bypassed by exploiting role confusion, study finds
A new paper titled "Prompt Injection as Role Confusion" by Charles Ye, Jasmine Cui, and Dylan Hadfield-Menell explores a vulnerability in large language models (LLMs) where safety rules can be bypassed through role impe…
-
Prompt injection exploits LLM role confusion, new research finds · 8 sources tracked
New research indicates that prompt injection attacks exploit a fundamental flaw in how large language models perceive roles, rather than a lack of safety filters. Researchers found that models prioritize the stylistic p…