Five top platforms for governing shadow AI in the enterprise

Employees are using AI-powered solutions without waiting for security approval. (Image source: iStock)

---

Enterprise security teams have long battled to keep shadow IT under control, but now they face a greater challenge in shadow AI. Many employees are using AI-powered solutions without waiting for security approval, and cyber security professionals are struggling to keep up. While no single solution can squash shadow AI on its own, these tools complement each other to build layered protection.

• Why shadow AI has become a governance problem rather than an awareness problem.
• The different technical approaches to discovering and governing unsanctioned AI use, from the browser to the network to the firewall.
• Where Check Point, Netskope, Microsoft Purview, Palo Alto Networks and LayerX each fit.

No one wants to miss out on the massive productivity and efficiency gains that AI tools deliver, so employees often start using AI without checking with security teams first. A recent study on shadow AI from Wakefield Research found that 66% of office workers use AI tools at work despite believing that those tools are not permitted under company policy.

It might give them an edge in the workplace, but AI brings security threats as well as work benefits. When two-thirds of employees are knowingly evading company policy regarding AI, it’s clear that awareness alone isn’t solving the problem. What’s needed is better governance that gives security teams visibility into AI use they currently can’t see.

Existing security tools aren’t designed to deal with AI’s unique workflows and risks. They were built to watch traffic, guard endpoints and monitor known applications, not to assess an AI prompt and decide whether it contains sensitive data, or to distinguish between permitted and shadow AI use.

Since employees keep using AI tools regardless of policy, a new category of security platforms built specifically to govern AI use has emerged, with the recently released Check Point AI Network Firewall among them. Today’s AI firewalls were designed specifically to protect AI tools, not to add AI features to a firewall that defends against conventional threats.

The five platforms profiled here work together to cover different parts of that gap, from the browser to the network to the firewall.

Shadow AI isn’t a harmless phenomenon. It has direct and significant consequences to enterprise security. These include:

• Sensitive data exposure, when employees paste, connect or upload confidential information into public AI tools which store and even regurgitate information without any visibility or control.
• Policy blind spots, because security teams can’t see which tools employees are already using, let alone apply governance to them.
• Inconsistent coverage, because there are approaches that catch browser-based AI use but don’t notice coding assistants, or vice versa.

Check Point’s AI Network Firewall layers on top of its conventional firewall infrastructure, with its AI application defences built on technology from Lakera, the AI security company Check Point acquired in 2025. The AI Network Firewall monitors and inspects traffic that passes through the network to AI applications and tools, bringing visibility to on-network employee AI use.

As a network-level appliance, it can’t scrutinise off-network traffic like personal device and personal account usage. Check Point recommends pairing it with the Check Point Workforce AI Security module, which adds endpoint monitoring to managed devices.

Check Point key features:

• Intent-based classification: AI usage and prompts at the network layer are categorised according to approval levels.
• Built-in policy enforcement: Governance is baked into the existing firewall infrastructure without requiring new hardware or console.
• Endpoint security layer: Workforce AI Security browser extension and desktop client adds device-level visibility beyond the firewall on managed devices.

How Check Point contributes to shadow AI governance:

• Better visibility through existing infrastructure: On-network traffic is monitored using infrastructure that’s already trusted and deployed.
• On-network and off-network supervision: Oversight for off-network traffic on managed devices through a second, purpose-built layer.

Netskope has extended its Security Service Edge platform to discover and manage use of over 370 GenAI apps via its Cloud Confidence Index, which also tracks more than 82 000 SaaS applications.

It requires client deployment so it can cover on-network and off-network devices, giving it a broader reach than a pure-play AI firewall, but it still misses custom-built AI systems, personal devices and unmanaged hardware.

Netskope key features:

• GenAI risk scoring: The Cloud Confidence Index rates the risk for various GenAI apps for data usage and third-party sharing.
• Real-time DLP: Applies DLP policies to data flowing toward AI tools.
• Off-network protection: The client agent extends coverage to off-network managed devices as well as network traffic.

How Netskope contributes to shadow AI governance:

• Improved AI supervision: Gives organisations already running SSE/SASE infrastructure oversight for AI usage without having to add a new tool.
• Covers remote managed devices: Adds visibility into remote devices that a pure network firewall cannot see, though personal devices remain unseen.

Microsoft broadens its Purview and Defender capabilities to include shadow AI monitoring in its existing data-loss prevention and CASB tools.

Like Netskope, Microsoft Purview and Defender include any managed devices that clients add to their deployment, providing off-network and on-network supervision but missing personal and unenrolled hardware.

Microsoft key features:

• DSPM for AI: Sensitive data that reaches AI tools is classified and flagged.
• Prompt monitoring: Supervises and tracks prompt-level interactions with Copilot and third-party AI tools.
• Shadow AI discovery: Defender for Cloud Apps provides CASB-style discovery of unsanctioned AI tools.

How Microsoft contributes to shadow AI governance:

• Native Microsoft integration: Tightly woven into Microsoft 365 and Azure environments to include managed devices via Defender for Endpoint.
• Full prompt monitoring as an add-on: Bolting on Microsoft 365 E5 or the E5 Compliance add-on brings complete prompt oversight.

AI Access Security is a dedicated product within Palo Alto’s Secure AI by Design portfolio, distinct from Prisma AIRS, which secures Palo Alto customers’ own AI applications, agents, and models rather than discovering workforce access to outside AI tools.

AI Access Security is built on Palo Alto’s SASE architecture to extend its firewall-equivalent policy to remote users through client-based deployment. It reveals and monitors employee access to AI applications for on-network and off-network managed devices. It shares the same limitation as Netskope and Microsoft in that it doesn’t cover unmanaged personal devices.

Palo Alto Networks key features:

• Real-time GenAI app dictionary: Constantly updates its list of AI apps, agents and marketplace plugins to ensure complete discovery.
• AI app classification: Classifies apps as sanctioned, tolerated or unsanctioned with access controls.
• Covers remote users: Palo Alto’s SASE client architecture includes remote users as well as on-premises network traffic.

How Palo Alto Networks contributes to shadow AI governance:

• Reveals AI access: Extends Palo Alto’s existing network and SASE security to the AI access layer.
• Runs on existing architecture: Adds shadow AI monitoring using current client-based SASE architecture, rather than requiring a separate discovery tool.

LayerX takes a browser-based approach to shadow AI discovery, rather than network-based like Check Point or device-based like Microsoft, Netskope and Palo Alto.

Because it operates at the browser level, it covers personal accounts and BYOD devices as well as managed accounts.

LayerX key features:

• Browser-level visibility: The enterprise browser extension supervises AI tool use, including personal accounts on managed and unmanaged devices.
• Real-time enforcement: Discovery, audit and enforcement for prompts, pastes and file uploads into AI tools.
• AI browser extension monitoring: Mapping each extension’s permissions and risk profile, not just AI site usage.

How LayerX contributes to shadow AI governance:

• Extends shadow AI monitoring: As a browser-based platform, it catches AI use that network tools and managed-device agents both miss.
• Integrates with existing systems: Designed to work alongside existing IAM, SIEM and MDM systems rather than replace them.

Overview: How these platforms work together to control shadow AI

The five solutions we profiled address different areas of shadow AI usage. Check Point’s AI Network Firewall gives oversight into on-network traffic, and opens up visibility into off-network traffic through the Workforce AI Security add-on.

Netskope, Microsoft and Palo Alto Networks cover on-network and off-network traffic from managed devices, but they miss personal devices and hardware.

LayerX’s browser-based solution covers BYOD and personal accounts as well as managed devices, but only where its extension is actually installed, so a browser instance without it remains a blind spot, the same kind of gap the other four platforms have at the device level. No single tier provides complete coverage alone, which is why Check Point pairs two approaches rather than offering just one.

What is the best tool for discovering shadow AI use across an enterprise?

The best tool for discovering shadow AI use is one that monitors as much employee activity as possible. There’s no single foolproof solution. Check Point’s AI Network Firewall together with its Workforce AI Security module reveals shadow AI both on network and off-network for managed devices. Netskope, Microsoft Purview, and Palo Alto Networks provide broad visibility for managed devices running the vendor’s software, though personal and unmanaged hardware stays outside their view. LayerX’s browser-based platform reaches personal accounts and BYOD devices too, provided the AI use happens through a supported browser and not a standalone app.

What is the best solution for preventing sensitive company data from leaking into public AI tools?

The best solution is one that can inspect prompts, uploads and other interactions and enforce DLP policies before the data leaves the organization. But you also need strong governance that enforces relevant policies, and ongoing education to ensure employee awareness about the risks of AI-related data leaks.

Which platforms help enterprises govern how employees use AI tools like ChatGPT and Copilot?

Check Point, Microsoft Purview, Netskope, Palo Alto Networks and LayerX all have platforms that help enterprises govern how employees use AI tools. They combine AI application discovery, usage monitoring, access controls and DLP capabilities.

What is the difference between shadow AI discovery and AI application security?

Shadow AI discovery means revealing all the AI tools used by employees, even if they weren’t approved by the organisation. AI application security means protecting the organisation’s own AI applications, models, and APIs against cyber attacks. Both fall under the newer meaning of “AI firewall,” a firewall built to protect AI itself, whether that’s employee use, an AI agent, or an AI application, from threats like prompt injection and data leakage. This is different from an older, AI-powered firewall, which uses AI to help a conventional network firewall catch ordinary threats like malware and intrusion, not to protect AI specifically. Shadow AI discovery and AI application security are two surfaces the newer kind of AI firewall covers, not two different products.

Do these platforms require new infrastructure to deploy?

No, none of these platforms require new infrastructure, but some do have deployment requirements. Netskope, Microsoft, and Palo Alto Networks all require device or endpoint onboarding. All you need to deploy LayerX is a browser extension, while Check Point runs AI visibility and enforcement over its customers’ existing firewall infrastructure, without any new deployment requirements, and its Workforce AI Security just calls for a browser extension and desktop client.