Tag: AI safety

An inductive analysis of 169 usable responses found that AI trust centers on performance while substantial shares also emphasize safety, ...

A peer-reviewed review proposes IDEAI, a framework for representing, checking and revising the intent that guides embodied AI actions in ...

A 2026 editorial argues that medically useful vision-language systems must join reliable multimodal perception to explainable, uncertainty-aware and safety-tested reasoning.

A single-run benchmark of 120 liver injury narratives found that AI systems produced parseable output while retaining potentially causality-distorting errors, ...

A new opinion article argues that AI agents could theoretically exploit human social media sharing as unwitting infrastructure for broadcasting ...

A new Academic Psychiatry commentary argues that clinicians and educators must distinguish evidence about AI models from evidence about deployed ...

A new study finds that large language models rewriting medical literature for patients retain only about half of the original ...

Researchers propose a five-plane software topology that separates decision, action, oversight, and adaptation to make tool-using AI agents structurally safe ...

Researchers have shown that concepts erased from text-to-image diffusion models can reappear when structural controls like edge maps and depth ...

A new Nature Machine Intelligence perspective argues that the reasoning tokens produced by large language models should be understood as ...