What happened
F5 launched Workforce AI Security, extending its existing AI Security Platform to provide visibility and control over employee interactions with AI services and the subsequent actions taken by AI agents. The product focuses on 'AI acting with borrowed authority,' where agents use user credentials to access internal systems, call tools, or modify data. It operates within the network path rather than requiring separate endpoint installations, allowing security teams to inspect, allow, block, or modify tool calls across MCP servers and various agent interfaces based on identity and risk policies.
F5 has launched Workforce AI Security, a new offering that extends its AI Security Platform to monitor how employees use artificial intelligence tools and to govern the actions of AI agents acting on their behalf. The product is designed to give organizations visibility into which AI services staff are using and allows security teams to set rules around those services, covering aspects such as license tiers, file uploads, and data policies.
The launch responds to the trend of companies giving AI systems wider roles in operations, where employees use AI assistants and coding agents that can access internal systems, call software tools, and carry out tasks with permissions granted by staff. F5 cites its State of Application Strategy Report, noting that 66 percent of organizations already allow AI to adjust policies and configurations automatically, raising questions about governing the actions these tools take once employees delegate work to them.
Workforce AI Security specifically targets what F5 describes as 'AI acting with borrowed authority,' where an agent carries out actions using a user's credentials or permissions. The system creates an audit trail showing who initiated an AI task, which agent executed it, what the system attempted to do, and whether the action was allowed or blocked. It identifies AI interactions by user and agent, recording context including intent, risk, and policy decisions for enforcement and audit purposes.
A key feature of the service is its ability to inspect actions before they are executed. The system checks tool calls across MCP servers and supported agent tools, then allows, blocks, or modifies those actions based on identity, access risk, and the exposure of sensitive data. F5 states that these controls are applied in the interaction path rather than through a separate endpoint installation, meaning the service works across browsers, command-line interfaces, coding agents, MCP clients, and internally built tools connecting to public model APIs.
This release builds on F5's broader push to expand its AI security portfolio, which earlier included the introduction of its AI Security Platform for visibility, governance, testing, and runtime protection, as well as AI Gateway functions like MCP Gateway. Workforce AI Security shifts part of that focus to employee activity and delegated agent behavior, aiming to provide a single policy layer that is not tied to one model, application, or vendor stack.
Source details: itbrief.asia ↗
Why it matters
This launch addresses a critical gap in enterprise security as AI shifts from passive text generation to active execution of tasks within business systems. By placing controls in the interaction path, F5 aims to provide a vendor-agnostic policy layer that can audit who initiated an AI task, which agent executed it, and whether the action complied with data policies. This is significant for organizations managing a mix of consumer AI services, coding assistants, and internal agent tools, as it moves oversight beyond simple content monitoring to the control of machine-led activity inside enterprise environments.
The product addresses a significant shift in enterprise security challenges as AI tools move beyond generating text into direct action inside enterprise environments. Once an employee allows an agent to retrieve data, interact with software tools, or change settings, oversight shifts from simple content monitoring to the control of machine-led activity inside business systems.
For companies, the issue is no longer only whether staff are entering sensitive information into AI tools, but whether an agent can trigger a change, access an internal resource, or carry out a workflow step without sufficient review. F5 aims to address this gap by combining discovery of AI tool usage with policy checks on agent behavior, giving security teams one place to manage both employee AI activity and actions performed on employees' behalf.
By placing controls where prompts and responses move across networks, F5 aims to offer a single policy layer that is not tied to one model, application, or vendor stack. This approach is intended to help organizations discover AI use, test for vulnerabilities, protect live interactions, and govern what users and agents can access without changing existing systems, fitting into existing SASE environments.
What to watch next
Organizations should monitor how this network-path approach integrates with existing SASE environments and whether it effectively captures traffic from diverse agent frameworks. The practical implication is a new layer of governance for delegated agent behavior, potentially reducing the risk of unauthorized data access or system changes initiated by AI tools. However, specific pricing, general availability dates, and detailed technical specifications for the MCP gateway integration were not provided in the source.
The practical implication of this launch is a new layer of governance for delegated agent behavior, which may reduce the risk of unauthorized data access or system changes initiated by AI tools. Security teams can now enforce intent-based guardrails in the network path to understand the context of AI interactions and enforce policy before risky actions occur.
Organizations should watch how this network-path approach integrates with existing SASE environments and whether it effectively captures traffic from diverse agent frameworks, including those using MCP servers. The source does not provide specific pricing, general availability dates, or detailed technical specifications for the MCP gateway integration, so these details remain unknown.
The launch highlights the growing need for enterprises to bring order to a mix of consumer AI services, coding assistants, and agent-based tools entering the workplace. As these systems expand the number of services and workflows security teams must track, the ability to audit and control agent actions becomes a critical component of enterprise AI strategy.