Major corporate insurance underwriters are rewriting their cybersecurity policies to account for risks from autonomous software agents. The changes follow several incidents where automated tools caused database damage or leaked data without human input.
Under the new terms, companies must prove their AI agents are sandboxed. They must also show that human supervisors can audit and override agent actions. Firms that cannot provide these documentation records face higher premiums or policy cancellations.
Managing Autonomous Software Risks
As businesses deploy agents for customer service and billing, the potential for error grows. A misconfigured agent can execute thousands of incorrect transactions in minutes before anyone notices the error.
"Standard cyber policies were designed for external hacks," says risk analyst Leila Nasseri. "Now, we are seeing claims from internal software errors. Underwriters are adjusting their risk models to ensure companies monitor their automated tools."
The new policy guidelines will become standard for all corporate renewals starting next quarter.
