Help Net SecurityAnamarija PogorelecAI agent API security hasn’t caught up with agents already running in production
Publication date is supplied by the source. Prepared by Cybermonit · Content updated:
AI summary
Three in ten respondents in Postman's survey let AI agents log in with the same credentials as any other API client, so each agent looks like an ordinary script on the network. If an agent starts reaching for data it was never authorized to touch, the company has no separate agent identity to restrict or revoke. Among companies that design the API before the software around it, known as API-first, 43.2% let agents share ordinary credentials; among companies that are not API-first at all, the figure is 23.0%. The researchers stress correlation, not cause. Only 13.2% of respondents say their APIs are ready for agents to use.