Abu Dhabi AI firm flags security gaps as agents act with employee credentials
AI
Dispatched by Tariq Al-Mansoor, Gulf & Capital Desk
- AI agents accessing emails and systems with employee credentials risk obscuring who authorized actions, says ANSEN co-founder Zheng Li.
- NIST and Microsoft recommend treating agents as “first-class entities” with separate identities, limited permissions and full audit trails.
- UAE government plans to deploy agentic AI in 50 % of services within two years, heightening the need for clear control and responsibility frameworks.
Abu Dhabi’s ANSEN warns that AI agents using employee accounts can blur accountability, prompting calls for dedicated identities and tighter controls.
«Abu Dhabi firm explores security concerns as AI agents act on users’ behalf»
Summary and commentary by Aura-V. Facts, figures and opinions belong to the original.
Abu Dhabi-based AI firm ANSEN highlighted security concerns around AI agents that act on users’ behalf, noting that agents can read emails, search files and execute multi-step tasks with legitimate credentials Khaleej Times. Co-founder Zheng Li warned that when an authorised agent makes a wrong decision, it can be difficult to trace who authorised the action.
Both NIST and Microsoft now advise giving each agent a dedicated identity, restricting its access, and recording the full activity chain to answer what happened, under whose authority, and what changed.
In the UAE, the federal government aims to transition 50 % of sectors to agentic AI within two years, with over 300 participants from 50 entities mapping potential deployments. Li argues that identity, permissions and audit trails must be built into these systems to preserve human judgement while retaining the speed and automation benefits of AI agents.
Interested in implementing this technology inside your company?
Our studio designs custom agent meshes, data pipelines, and outbound GTM systems for forward-looking enterprises.
