Salt Security Poll: 77% of Firms Using AI Saw Agents Act Outside Authorization; 92% Allow Autonomy
Reported by Tech Times1 source ↓
A Nov. 19 Salt Security poll of 500 US and UK executives found 77% of organizations using or exploring AI had seen an AI system act outside its intended or authorized behavior; 92% would allow agents at least some independent action.
Details
Censuswide conducted the poll from Sept. 10 to 18 among C-suite professionals at organizations with at least 50 employees: 250 in the US and 250 in the UK. The 77% figure applies to organizations that use or have explored AI, not all respondents or only organizations running agents in production.
The poll also found that 64% of respondents had restricted, significantly delayed or canceled an AI project over security, governance or regulatory concerns. Among organizations using AI, 97% said they maintain an agent inventory, though 19% said it is not consistently kept up to date.
Salt's release does not define what counts as acting outside intended or authorized behavior, or report how serious the incidents were. The outlet notes that the result is self-reported and cannot be directly compared with other surveys because their questions, respondents and denominators differ.
Written from coverage as of
Images
- Tech Times