AI Safety, Risk Perception & Boundary Behaviour 2025 (Dataset, n = 301)

This dataset is part of the Human Clarity Institute’s <b>AI–Human Experience 2025</b> data series. It examines AI safety perceptions, risk boundaries, comfort levels, and responsible-use expectations, including risk tolerance, boundary-setting behaviours, and trust in AI across everyday contexts.The dataset includes:<br>• validated 1–7 Likert-scale items<br>• measures of AI safety perception, perceived risk, personal comfort, and responsible-use expectations<br>• behavioural indicators related to boundary-setting, acceptable AI use, and scenarios that feel unsafe or concerning<br>• multi-select variables stored as canonical semicolon-delimited snake_case tokens<br>• open-text reflections with minimal safe cleaning (trim + newline removal only)<br>• demographic variables across six English-speaking countries<br>• digital life exposure (daily hours online) and AI-tool usage frequencyData were collected on 1 December 2025 via Prolific from adults in the UK, US, Australia, Canada, New Zealand, and Ireland.<br>All data were cleaned, anonymised, and processed under the Human Clarity Institute’s machine-readable dataset protocol, which includes:• canonical snake_case variable naming<br>• validated numeric ranges<br>• standardised multi-select formats<br>• minimal safe text cleaning<br>• full alignment with the accompanying data dictionary<br>• removal of Prolific IDs and timestamps<br>• SHA-256 checksums for all filesThis dataset contributes to understanding how people perceive AI safety and define acceptable boundaries for AI behaviour, supporting longitudinal tracking of shifts in risk tolerance, comfort, and trust and the evolving cognitive relationship between humans and AI in everyday life.