13:05Machine-translated from Azerbaijani by AIRead in Azerbaijani
53% of UK IT Leaders Say an AI Program Has Broken the Rules
In the UK, 53% of executives responsible for information technology said that at least 1 artificial intelligence (AI) agent had violated company rules. These violations affected the business or its customers. An AI agent is a program that performs assigned tasks with a certain degree of independence. The average in other countries was just 31%. The large disparity indicates problems with the oversight of AI systems in the UK.
“Harris Poll” conducted the research for “Dataiku.” The survey covered 685 CIOs working in the UK, France, Germany, the US, the UAE, Japan, South Korea and Singapore. A CIO is the senior executive responsible for a company’s information technology. Only 49% of the British participants said they could provide audit logs explaining the AI system’s decisions. According to 68%, detecting an agent that has gone out of control and limiting its impact could take more than 24 hours. 87% of British IT leaders identified the rapid proliferation of specialized AI agents in workplaces as the main problem. “Dataiku” calls this situation “AI agent sprawl.” Such agents are deployed before oversight rules are developed. As a result, they can create additional risks for both the company’s operations and its customers.
According to the research, executives expected to deliver results from AI investments are under intense pressure. Most participants believe budgets could be reduced and jobs could be cut if they fail to meet their targets. Although companies monitor the agents’ work, they struggle to identify the responsible person when a problem arises. According to “Dataiku” executives, simply monitoring systems that operate independently is not enough. They must be kept under constant supervision and managed. If an agent violates a rule, responsibility must be established before the system is deployed. Adding oversight after a problem occurs is not considered sufficient. Otherwise, the company’s only option is to stop the agent and hope that the consequences do not become more severe.
“Harris Poll” conducted the research for “Dataiku.” The survey covered 685 CIOs working in the UK, France, Germany, the US, the UAE, Japan, South Korea and Singapore. A CIO is the senior executive responsible for a company’s information technology. Only 49% of the British participants said they could provide audit logs explaining the AI system’s decisions. According to 68%, detecting an agent that has gone out of control and limiting its impact could take more than 24 hours. 87% of British IT leaders identified the rapid proliferation of specialized AI agents in workplaces as the main problem. “Dataiku” calls this situation “AI agent sprawl.” Such agents are deployed before oversight rules are developed. As a result, they can create additional risks for both the company’s operations and its customers. According to the research, executives expected to deliver results from AI investments are under intense pressure. Most participants believe budgets could be reduced and jobs could be cut if they fail to meet their targets. Although companies monitor the agents’ work, they struggle to identify the responsible person when a problem arises. According to “Dataiku” executives, simply monitoring systems that operate independently is not enough. They must be kept under constant supervision and managed. If an agent violates a rule, responsibility must be established before the system is deployed. Adding oversight after a problem occurs is not considered sufficient. Otherwise, the company’s only option is to stop the agent and hope that the consequences do not become more severe.