A survey released by the British Standards Institution (BSI) found that 74% of Japanese companies had experienced an AI incident in the past year, highlighting the state of their organisational frameworks and the impact of incidents. Incidents involving agentic AI are estimated to put up to 3.6% of market revenue at risk.
According to a survey conducted by BSI, 1,287 business leaders in 9 countries, including Japan, were asked about their experience of AI incidents and their impact. In Japan, 74% of companies had experienced an incident in the past year, while 91% said that incidents had occurred multiple times.
The reported impacts included effects on safety at 38%, regulatory and compliance issues at 35%, security breaches at 31% and wasted working hours at 30%. Only 14% of companies reported financial losses. Just 41% had established processes for managing AI incidents.
Economic modelling estimated that AI incidents could put up to 3.6% of market revenue at risk annually. By type, agentic AI accounted for up to 3.6%, hallucinations and similar issues for 2.5%, biased decision-making for 0.3% and customer-service chatbots for 0.1%.
Only 13% of Japanese companies had moved agentic AI into full-scale operation. Security was a concern for 63%, the lack of AI governance for 42% and inadequate regulation for 30%. A total of 73% expected to increase their AI investment.
Companies with an AI governance programme accounted for 40%. Those that had properly implemented policies for safe AI use in compliance with standards accounted for 35%, while 34% had secured transparency for data and algorithms.
BSI Group Japan said it would support companies through guidance and certification based on ISO/IEC 42001.