AI agents can carry out tasks independent of human oversight, but businesses need to decide how much access to give them. If a tool makes a mistake, access limits can help keep the problem from spreading to other company systems.
These limits define which systems an agent can use and where it can send information. For companies putting artificial intelligence (AI) to work, the question is not just whether it can finish a task, but whether it can do so without reaching files or services outside its assigned work.
“With iboss, we can set policies, even for agentic AI or autonomous agents,” said Joseph Angelo Tadaya, senior solution engineer and senior sales manager for Asia Pacific, iboss, a Zero Trust Secure Access Service Edge (SASE) cloud security platform.
In an interview on the sidelines of VSTECS CXO Tech Summit held in Boracay, Oct. 20-23, Tadaya said businesses can restrict an AI agent’s connections to an approved list. This means setting rules on which systems and websites it can reach and what it can communicate with.
The approach gives companies a way to limit where an agent operates, particularly when it works with little supervision. Tadaya said these restrictions help protect sensitive information by preventing connections to destinations outside the list.
Access rules are one part of managing AI use at work. Businesses also need to consider what employees share with AI tools while asking for help with everyday tasks.
Tadaya cited employee records and health records as examples of sensitive information that could leave company systems through AI conversations. He said iboss can check those conversations for sensitive content and block uploads.
Such controls are intended to let employees use AI for tasks including writing code, checking code, and troubleshooting, while setting boundaries on the information they can submit.
iboss provides a software-as-a-service solution that provides security and networking to give users secure connections in the office or remotely.
iboss entered the Philippine market in 2021 and has since served customers in banking, hospitality, and government.
For businesses that already have cybersecurity tools, the company offers standalone AI security. This is an option for adding protection for AI use alongside an existing security provider.
The company supports cloud, on-premises, and hybrid setups, depending on customer requirements. Some banking customers still prefer security equipment within their offices, he said.
For businesses, the starting point is deciding what each AI agent needs to reach. A task should come with clear access boundaries, so allowing a tool to help with work does not mean opening every company system to it.