RealityTest: Do AI systems disclose their identity when asked?
Introduces RealityTest, a multilingual, text and speech benchmark built from real user identity probes, to test whether AI systems honestly disclose being AI when asked.
Research / Sources
UK AI Security Institute
The UK government’s AI Security Institute: evaluations, red-teaming and research on frontier AI risks. aisi.gov.uk
Organisation (36)Red Team (13)Science of Evaluations (11)Cyber & Autonomous Systems (10)Engineering (6)Human Influence (5)Control (4)Societal Resilience (4)Safety Cases (3)Chem-Bio (2)Model Transparency (2)Strategic Awareness (1)
Focus All
Topic All
Date Any time
Ask Don't Tell: Reducing Sycophancy in Large Language Models
Controlled study showing user input framing, perspective and professed certainty measurably change LLM sycophancy, and proposes a prompt-phrasing strategy to reduce it.
How do AI models persuade? Exploring the levers of AI-enabled persuasion through large-scale experiments
Large-scale experiments (76,000+ participants, 19 models, 700+ political issues) measure what makes conversational AI persuasive, finding levers like model scale, post-training, and rhetorical strategy.
Do chatbots inform or misinform voters?
UK study of voters found about one in eight used AI chatbots for political information, and chatbot users became no less informed than search users.
Should AI systems behave like people?
Surveys public preferences on whether AI should behave like people, motivated by risks of anthropomorphic AI such as impersonation, deception and emotional over-attachment.