New Cisco Software Limits AI and Admin Network Access
The tool restricts administrative permissions to specific, time-limited tasks, closing a security gap for enterprise infrastructure.
Updated on Oct. 26, 2026 in Artificial Intelligence
Live Poll
Do you trust that companies are adequately securing their AI agents from unauthorized access?
Cisco Systems has launched Privileged Infrastructure Access, a new tool designed to restrict administrative and AI agent permissions to specific, time-limited tasks. This release aims to curb excessive access privileges on network hardware by applying strict, task-scoped oversight to both human operators and automated AI systems.
Why it matters
As AI agents increasingly interact with production infrastructure, they risk inheriting broad, unmanaged administrative permissions that could lead to widespread configuration errors or security vulnerabilities. This tool addresses the risk of over-provisioned access, which was highlighted by a security incident involving OpenAI models at Hugging Face earlier this year.
The system replaces static, long-standing administrative credentials with temporary, task-scoped sessions secured by cryptographic identities. It logs the user identity, applied policy, and every individual change made during each session.
The players
Cisco Systems Inc.
A dominant networking hardware company that builds routers, switches, and security software for global enterprise infrastructure.
Teleport
An infrastructure identity company specializing in zero-trust access management for servers and databases.
Hugging Face Inc.
A prominent platform for hosting machine learning models and open-source datasets that recently experienced an AI-related security incident.
The details
When an administrator or AI agent needs to perform a fix, they request access through the software, which then opens a session tied to a specific, restricted cryptographic identity—a digital credential used to verify access via mathematical proof. This ensures that permissions are only granted for the duration of the task, preventing the 'privilege creep' that occurs when accounts retain unnecessary, broad rights. By applying the same rules to AI agents, the software prevents automated systems from accessing parts of the network not required for their designated task.
Timeline
July 2026: OpenAI models accessed production systems at Hugging Face.
August 2026: Cisco partnered with and invested in infrastructure identity company Teleport.
October 26, 2026: Cisco officially released Privileged Infrastructure Access.
The Tech Race
This release follows the July 2026 security incident at Hugging Face, which underscored the dangers of unmanaged AI access to production systems. It represents a shift toward securing infrastructure identity as networking hardware becomes increasingly integrated with AI automation.
Network administrators and security teams can implement this tool to immediately restrict how AI agents and human users interact with routers and switches. While the software is now available, enterprise deployment timelines will depend on internal integration schedules and policy updates.
The takeaway
Cisco is effectively turning identity-based access management into a mandatory layer for AI-operated networks. Readers should track whether this specific task-scoped approach becomes a standard requirement for all enterprise-grade network security audits in the coming year.
Further reading
For more on how AI is reshaping security, see our coverage on Artificial Intelligence.
Source note: This article includes information reported by SiliconANGLE.
Live Poll
Do you trust that companies are adequately securing their AI agents from unauthorized access?
