Survey: Consumers Wary of Unsupervised AI Agents, Citing Identity and Permission Gaps
Okta's 2026 Customer Identity Trends Report finds that only 37% of North American consumers trust AI agents to make final decisions or handle routine tasks autonomously, with willingness rising above 50% when users retain approval or reversal control. The report highlights that most organizations deploying AI agents face delays due to engineering capacity shortages and identity infrastructure unable to support agent-based access, with only 22.1% of initiatives progressing at the intended pace. The piece is a business/industry research summary rather than a security incident report, focused on identity and access management challenges around AI agent adoption.
Detalhes técnicos
The report notes that engineering teams spend roughly 14% of their time building and maintaining customer identity and access management (CIAM) systems. It recommends giving AI agents distinct identities with scoped, time-limited credentials restricted to required tasks, and maintaining activity logs to investigate errors or unauthorized actions. Shared accounts and embedded credentials are flagged as risks that could expose connected services if an agent is compromised. Organizations relying on external customer identity platforms reportedly show better on-schedule deployment and stronger authentication practices, though the report notes this is only a correlation, not causation.
Remediação
The report suggests organizations adopt per-agent identities with expiring, least-privilege credentials, implement robust audit logging of agent actions, and use dedicated customer identity platforms to centralize permission management and reduce duplicated engineering effort. Clear communication of permissions and easy mechanisms for customers to revoke or adjust agent access are also recommended.
Linha do tempo
-
Survey: Consumers Wary of Unsupervised AI Agents, Citing Identity and Permission Gaps
ler a análise desta fonte ocultar
Okta's 2026 Customer Identity Trends Report finds that only 37% of North American consumers trust AI agents to make final decisions or handle routine tasks autonomously, with willingness rising above 50% when users retain approval or reversal control. The report highlights that most organizations deploying AI agents face delays due to engineering capacity shortages and identity infrastructure unable to support agent-based access, with only 22.1% of initiatives progressing at the intended pace. The piece is a business/industry research summary rather than a security incident report, focused on identity and access management challenges around AI agent adoption.
Technical Details
The report notes that engineering teams spend roughly 14% of their time building and maintaining customer identity and access management (CIAM) systems. It recommends giving AI agents distinct identities with scoped, time-limited credentials restricted to required tasks, and maintaining activity logs to investigate errors or unauthorized actions. Shared accounts and embedded credentials are flagged as risks that could expose connected services if an agent is compromised. Organizations relying on external customer identity platforms reportedly show better on-schedule deployment and stronger authentication practices, though the report notes this is only a correlation, not causation.
Remediation
The report suggests organizations adopt per-agent identities with expiring, least-privilege credentials, implement robust audit logging of agent actions, and use dedicated customer identity platforms to centralize permission management and reduce duplicated engineering effort. Clear communication of permissions and easy mechanisms for customers to revoke or adjust agent access are also recommended.