Companies Have a New Security Problem: AI Agents They Don't Know Exist

Gartner warns that employees are creating and using AI agents faster than many organisations can govern them.

Gartner reports that 75% of organisations see unauthorised use of AI coding assistants, while 50% report employees accessing public SaaS GenAI-agent platforms.

These "shadow AI" systems can interact with organisational information and services without security teams having sufficient visibility or control.

For AI engineers, this reinforces the importance of:

Permissions + Authentication + Access Control + Security + Governance + Monitoring

As AI systems gain more autonomy, understanding what they are allowed to access and do becomes part of engineering the system itself.

Takeaway: Building AI agents without understanding permissions and security creates real organisational risk.