How Check Point Workforce AI Security works

What Workforce AI Security is for

Check Point Workforce AI Security is Check Point's product for securing the way employees use AI. Check Point sums it up as a way to "discover, govern, and protect employee AI usage across browser, desktop, and agents", and its administration guide describes a platform that gives visibility of approved and shadow AI tools, evaluates user sessions for risk, and can block risky activity and prevent data leakage.

That covers the AI people actually use at work: chat assistants such as ChatGPT, Gemini, Claude and Copilot in the browser, AI features inside SaaS apps, desktop AI apps, coding assistants in developers' editors, and AI agents connected to company systems through the Model Context Protocol (MCP). It is one of the modules of Check Point's AI Defense Plane, announced in March 2026, alongside the products for AI applications and agents that organisations build themselves (see Check Point AI Security).

This page goes under the bonnet. If you want to know which tier you need, jump to Workforce AI Security Essentials, Workforce AI Security Enterprise or the side-by-side Essentials vs Enterprise guide.

The architecture in three layers

Check Point's administration guide describes three layers:

• Endpoint monitoring layer: components on each device that capture AI-related actions and send them to the cloud for analysis. These are a browser extension for web-based chats and tools, a desktop client for desktop AI apps and coding assistants, integrations for developer tools with AI features, and visibility of MCP servers used by agentic workflows.
• Policy and classification engines: cloud services that evaluate each interaction for risk, sensitive data and use case, then apply your Access, Chats (DLP) and Agent policies.
• Management and governance layer: the Check Point Portal, where administrators see dashboards, discovered apps and agents, event logs and reports, and configure policy, with role-based access that controls who can see sensitive prompts.

Check Point sets out the data flow like this:

1. A user interacts with an AI tool: a web chat, a desktop app, an IDE assistant or an MCP agent.
2. The endpoint components capture the interaction (prompts, file uploads, pasted text, agent tool calls), subject to the permissions and privacy settings in force.
3. The cloud analysis engine evaluates risk, detects sensitive data and classifies the use case.
4. Policy rules are applied from the Access, Chats (DLP) or Agents policies.
5. An action is enforced: Allow, Ask, Block, Detect or Prevent.
6. The event is logged and appears in dashboards, so administrators can adjust governance.

Stage 1: Discover

Discovery answers the first question every organisation has: what AI is actually in use? Workforce AI Security shows sanctioned and shadow AI applications, the users and sessions behind each one, risk classifications, and recent events such as prompts and file uploads. A dedicated Agents view lists AI agents and MCP servers.

• Web AI apps: chat interfaces and AI tools used in the browser
• Desktop and developer tools: Check Point's guide names ChatGPT Desktop, Claude Code, GitHub Copilot, and editors such as Cursor and VS Code that use AI coding assistance and MCP servers
• Application catalogue: a searchable catalogue of AI applications, including ones nobody in your organisation has used yet, so you can set policy before adoption
• Inventory: a read-only view of AI agents and AI-enabled applications found on devices by Check Point's endpoint scanner, used to set a baseline before enforcement
• SaaS and platform integrations: connections to some enterprise AI platforms, such as Microsoft 365 Copilot, OpenAI's enterprise offering and Microsoft Copilot Studio, to discover agents and AI activity inside them

Our guide to shadow AI discovery explains why this step comes before any rule-writing.

Stage 2: Assess

Once you can see the tools, the next question is whether they are acceptable. Workforce AI Security helps in three ways:

• It classifies applications and sessions by risk level (critical, high, medium, low and so on), based on security posture and observed activity.
• It uses AI-based classification to sort prompt content, including files, into what Check Point calls "tens of use case categories", such as Marketing, Debugging, Legal and Email & Communication. That tells you what people use AI for, which is often the best guide to which tool you should approve.
• It distinguishes managed applications (used under an organisational licence) from unmanaged ones (personal or external accounts), so you can treat a corporate ChatGPT workspace differently from someone's personal login to the same service.

This is the evidence an AI risk assessment or an AI DPIA needs.

Stage 3: Govern

Governance in Workforce AI Security is a layered set of policies, each answering a different question:

Check Point names the policy actions as Allow, Ask, Block, Detect (log only) and Prevent. Policies can be set by application, user and data type, and the solution brief adds per-app copy and paste restrictions, file-based policies, prompt restrictions and rules against risky connections between AI tools and corporate resources. Users and groups come from your identity provider, so policies can follow your existing structure.

The policies are where your AI acceptable use policy becomes something that is enforced rather than just published.

Stage 4: Protect

Protection happens at the moment someone sends a prompt, uploads a file or lets an agent take an action:

• Contextual, AI-powered DLP: Check Point describes detection that goes beyond keywords, finding personal data, credentials, source code and financial data in prompts, uploads and attachments. Administrators can use predefined data types or build custom ones from patterns, keywords, dictionaries, weighted words, templates and file attributes.
• Inline redaction: sensitive values can be replaced with labelled placeholders (for credentials, personal data and more), so the user still gets an answer without the data leaving.
• File and image redaction with OCR: text inside images and documents is read and handled too.
• Copy and paste restrictions: limits on pasting sensitive content into particular apps.
• User guidance: when a policy blocks or prevents an interaction, the user sees why, including which data types triggered it, and Check Point's "action validation wizard" guides them to a safe way to continue.
• Agent guardrails: runtime controls that block risky agentic actions, such as a delete operation through an MCP tool.
• Prompt attack detection: Check Point's administration guide documents a Prompt Attack feature that analyses MCP tool definitions and tool call inputs and outputs for signs of prompt injection. When we checked, it was in Early Availability and ran in detect-only mode, recording events rather than blocking, so we confirm its status when we quote.

See AI data leakage for the risks this addresses.

Where it enforces: extension, desktop agent, firewall and SASE

Workforce AI Security can see and act in several places. Which you use depends on the tier and on how your people use AI.

If you already run Check Point firewalls, read our page on the Check Point AI Network Firewall. If you use Check Point's SASE service, our sister site covers Check Point Harmony SASE. Check Point notes that the options on the deployment downloads page depend on your licence, which is one of the things we confirm when we quote.

Audit trails, reports and integrations

Security and compliance teams need evidence, not just enforcement. Workforce AI Security logs events for each interaction and provides reports on AI usage, agent activity, detected data types, risks and enforcement actions over a chosen period. Check Point says its audit trails and customisable reports support GDPR, HIPAA and EU AI Act compliance work; they are evidence for your own assessments, not a compliance certificate. See AI and UK GDPR and does the EU AI Act apply to UK businesses.

For larger teams, Check Point documents public REST APIs for SIEM, SOAR and reporting integrations, a Terraform provider for managing policy as code, and an MCP server for automated operations. You do not need any of them to run the product day to day.

Rolling it out: a staged approach

Check Point's onboarding wizard runs through five steps (Welcome, AI Platforms, Manage Interactions, Deployment and Go Live), and its recommended defaults favour visibility first: Application Discovery for access, Monitor & Discover for chats, and Agents Discovery for MCP. We follow the same logic:

1. Pilot. Pick a representative group, including a few heavy AI users and, if relevant, some developers. Register your managed AI platforms so corporate and personal accounts can be told apart.
2. Discovery. Deploy the browser extension (and, for Enterprise, the desktop agent) in monitoring mode. Optionally run the read-only Agentic Endpoint Discovery scan for an early inventory.
3. Policy. Review what was found with the business, decide what to approve, restrict or block, and turn those decisions into Access, Chats and Agent policies, starting with warnings and redaction before hard blocks.
4. Roll out. Extend to everyone through your device management tools (Check Point documents Microsoft Intune and other platforms), then review the reports regularly and tune.

Rushing to block tends to push people to personal phones and workarounds. A staged rollout keeps AI useful while the risk comes down.

How Forever Systems helps

We help you get from "we think people are using AI" to clear evidence and sensible controls:

• We scope the work with you: who uses AI, in which tools, and whether desktop apps, coding assistants or agents are involved, which decides between Essentials and Enterprise.
• We help set up a pilot and read the discovery results with you.
• We help translate your AI policy into Workforce AI Security policies and plan the rollout.
• We quote for the licences you need. See how Workforce AI Security is licensed, and if you already have GenAI Protect, read our GenAI Protect end of sale guide.

For a vendor-neutral view of the wider market, our sister site compares approaches to AI usage control. To talk about Workforce AI Security for your organisation, contact us.