GlobalProtect VPN Explained: How It Works, Why It Drains Laptops, and When an On-Device SWG Is the Better Fit (2026)

GlobalProtect is Palo Alto Networks' VPN and endpoint agent. It builds an encrypted tunnel from a laptop or phone to a Palo Alto gateway (a next-generation firewall or Prisma Access) and sends the device's traffic through it so the gateway can inspect and enforce policy. That sentence contains the whole tradeoff: GlobalProtect is a tunnel client first and a security control second, because nothing is enforced until traffic reaches the gateway. Its documented battery drain, CPU spikes and reconnect loops are the cost of keeping that tunnel up. dope.security removes the tunnel instead of tuning it, by running Secure Web Gateway, CASB and DLP inspection on the device and sending traffic straight to its destination.

This explainer is for the IT or security lead who inherited GlobalProtect, is evaluating Prisma Access, or keeps getting tickets that start with "the VPN." It covers what GlobalProtect does, how it works, the configuration choices that matter, what Palo Alto's own documentation says about its failure modes, and when an on-device SWG is the better tool for the job. The single most consequential GlobalProtect decision is tunnel scope, so if you have not read split tunnel vs full tunnel VPN, start there and come back.

What is GlobalProtect?

GlobalProtect is the client software Palo Alto Networks ships for remote access. It runs on Windows, macOS, Linux, iOS, Android and ChromeOS, and it connects to two kinds of back end: a GlobalProtect gateway hosted on a Palo Alto next-generation firewall you operate, or Prisma Access, Palo Alto's cloud-delivered SASE service. In both cases the app authenticates the user, checks the device's posture with a Host Information Profile (HIP), establishes an IPsec or SSL tunnel, and steers traffic into it.

Because Palo Alto sells GlobalProtect as the on-ramp to both its firewalls and Prisma Access, the name gets used loosely. "GlobalProtect VPN" usually means the classic remote-access use case (get onto the corporate network from home). "GlobalProtect for Prisma Access" means the same agent used to backhaul all traffic, including internet-bound traffic, to a Palo Alto cloud location for SWG-style inspection. The second use is where it competes directly with dope.SWG.

How does GlobalProtect work?

The flow has four steps. The app contacts a GlobalProtect portal, which authenticates the user (typically through your identity provider via SAML) and hands back the client configuration and gateway list. The app then picks a gateway, usually the closest by measured latency, and negotiates a tunnel. Traffic that matches the tunnel configuration is encrypted and sent to the gateway. The gateway decrypts it, applies firewall, URL filtering, threat prevention and DLP policy, and forwards it on.

Three configuration modes govern how often that happens. Always-on keeps the tunnel up whenever the device has network. On-demand lets the user connect when needed. Pre-logon brings the tunnel up before the user signs in so domain scripts and MDM can reach the machine. Layered on top is the split-tunnel choice: send everything through the gateway (full tunnel) or exclude trusted destinations such as Microsoft 365 and Zoom (split tunnel) to save the gateway from carrying video traffic. Every one of these settings is a trade between control and user experience, and the trade exists because control lives at the far end of the tunnel.

The documented problems: battery, CPU and reconnects

The complaints about GlobalProtect are unusually well sourced because Palo Alto has documented several of them itself.

• macOS battery drain and CPU. Palo Alto has confirmed [Documented] GlobalProtect app versions that consumed excessive battery and pegged CPU at 100 percent on macOS, with fixes tracked across releases. Users on laptops notice this first, and it is the reason "GlobalProtect battery" is a support-forum staple.
• Reconnect loops. When a device moves between networks, wakes from sleep, or hits a captive portal, the app tears down and rebuilds the tunnel. Palo Alto documents [Documented] reconnect loops in specific version and gateway combinations, and customers report [Sentiment] that the app can sit in "connecting" for a minute or more during transitions.
• Setup complexity. Portal, gateway, HIP profiles, split-tunnel rules, certificate profiles and authentication sequences all have to agree. Setup complexity is the most common theme in Prisma Access reviews [Documented+Sentiment], and the Panorama-to-Strata Cloud Manager migration Palo Alto now recommends is one-way [Documented].
• Explicit Proxy limits. Where teams try to avoid the tunnel by using Prisma Access Explicit Proxy instead, they hit hard limits Palo Alto lists in its own docs: no HTTP/2, ALPN stripped, decryption mandatory [Documented].

One important clarification: the widely reported CVE-2024-3400 affected the GlobalProtect feature on on-premises PAN-OS firewalls, not Prisma Access. It is relevant to a firewall-hosted GlobalProtect deployment and should not be read as a Prisma Access issue.

Why the tunnel is the problem, not the app

Every issue in the previous section is a symptom of one design choice. GlobalProtect must keep a tunnel alive because the inspection engine is somewhere else. Keeping a tunnel alive means monitoring the network, re-authenticating, re-keying and re-routing every time the device's connectivity changes, which is constantly on a laptop that moves between home Wi-Fi, a phone hotspot and an office. The battery and CPU cost is the cost of doing that work. The reconnect loop is what it looks like when that work fails.

The agent-based alternative changes where the engine lives. dope.endpoint performs SSL inspection, URL filtering, Cloud Application Control and Dopamine DLP on the device, in under 100 MB of RAM. There is no tunnel to keep alive, so a network change is just a network change. Policy arrives from dope.console in seconds, enforcement continues from cached policy if the console is unreachable, and traffic goes from the device straight to the destination. The engineering comparison is laid out in Prisma Access vs an endpoint-native SWG, and the same pattern applies to the other steering agents in the category, as we showed with the Netskope Client and Zscaler Client Connector.

What does the GlobalProtect detour cost in latency?

When GlobalProtect is used for internet-bound traffic through Prisma Access, every request travels device to Prisma Access location to destination and back. Prisma Access runs on Google Cloud and AWS regions [Documented], and the added round trip depends entirely on how far the user is from the nearest one. Near a location the penalty is small; from a hotel in Manila or a client site in Perth it is not. Full-tunnel configurations make it worse by sending video and collaboration traffic on the same detour. We put numbers on this pattern in the cloud proxy latency math.

See what the detour costs on your own connection: the Fly-Direct Speed Test measures your real round-trip latency in the browser and projects what a cloud gateway hop adds to common apps. See how Fly Direct removes the hop, or book a 20-minute demo.

Takeaway: a gateway hop adds a fixed detour to every request; on-device inspection adds none, so the user's own connection is the load time.

GlobalProtect vs dope.security: paired comparison

These are the two designs side by side. Each bullet pairs the GlobalProtect behavior with the dope.security behavior.

• Where enforcement happens: GlobalProtect enforces at the gateway (firewall or Prisma Access) after the tunnel; dope.security enforces on the device in the dope.endpoint agent before traffic leaves.
• Network transitions: GlobalProtect tears down and rebuilds the tunnel, with documented reconnect loops; dope.security has no tunnel, so a network change does not interrupt enforcement.
• Resource footprint: GlobalProtect has vendor-confirmed macOS battery and CPU issues in specific releases; dope.endpoint runs in under 100 MB of RAM and delivers up to 4x performance over legacy proxy SWGs.
• Configuration surface: GlobalProtect needs portal, gateway, HIP, split-tunnel, certificate and auth profiles to agree; dope.security needs an MDM push and a policy in one console.
• Internet-bound inspection path: GlobalProtect through Prisma Access routes every web request to a cloud location and back; dope.security inspects locally and sends traffic direct.
• AI governance: Palo Alto's AI Access Security requires AI Access-X or CASB-X plus Enterprise DLP stacked on the decrypt proxy [Documented]; dope.security ships Shadow IT discovery, SWG policy and Cloud Application Control tenant restriction natively, with Dopamine DLP (US Patent 12,464,023) classifying prompts on the device.
• Control-plane risk: Strata Cloud Manager was impaired for roughly 28 days from March 31, 2026 [Documented]; dope.security enforcement is local and continues from cached policy during any console incident.
• China: Prisma Access in mainland China is partner-operated and requires ICP filing [Documented]; dope.security works in China with no data center to reach and no paid uplift.

When you still need GlobalProtect, and how to make it smaller

There is a legitimate job GlobalProtect does that an SWG does not: private application access to servers on your own network. If you run on-premises apps behind a Palo Alto firewall and are not ready to publish them differently, the VPN stays. What does not have to stay is the practice of pushing internet-bound traffic through the same tunnel. Many teams run GlobalProtect in on-demand or split-tunnel mode for private apps only and hand web, SaaS and AI traffic to an on-device SWG. We documented that coexistence pattern in running GlobalProtect, FortiClient and AnyConnect alongside dope.security. The result is a smaller, quieter VPN and a laptop that stops spinning its fan.

If the reason you are looking at GlobalProtect is Prisma Access as a full SASE replacement, read Palo Alto Prisma Access alternatives and Zscaler vs Prisma Access before you size the deal. Customers who moved web inspection to the endpoint, such as the healthcare organization in this displacement story, tend to describe the change in help-desk tickets before they describe it in security terms.

The bottom line on GlobalProtect

GlobalProtect is a competent tunnel client attached to a very capable gateway. Its problems are not bugs to be patched away; they are the recurring price of keeping a tunnel alive so that a distant engine can do the inspecting. Put the engine on the device and the tunnel, the battery drain, the reconnect loop and the detour all become optional. Start a free trial to run dope.SWG next to GlobalProtect on a handful of laptops, or book a 20-minute demo.

Frequently Asked Questions

Is GlobalProtect a VPN or a firewall?

GlobalProtect is the VPN client and endpoint agent. The firewall is the Palo Alto next-generation firewall or Prisma Access gateway it connects to. GlobalProtect itself does not inspect traffic; it authenticates the user, checks device posture and builds the tunnel so the gateway can inspect. dope.security is neither a VPN nor a gateway: it inspects on the device and needs no tunnel for web, SaaS and AI traffic.

Why does GlobalProtect drain my Mac's battery?

Palo Alto has confirmed releases of the GlobalProtect app that consumed excessive battery and CPU on macOS, and the always-on tunnel itself does continuous work monitoring the network and re-keying. Updating the app helps in the affected versions. Removing internet-bound traffic from the tunnel and inspecting it on the device with dope.SWG removes most of the work the app is doing.

Should I use split tunnel or full tunnel with GlobalProtect?

Full tunnel gives the gateway visibility into all traffic at the cost of routing video and SaaS through it. Split tunnel improves performance but creates blind spots for whatever is excluded. An on-device SWG resolves the dilemma: dope.security inspects the traffic you would otherwise exclude, so GlobalProtect can shrink to private-app access only. Our split tunnel vs full tunnel guide covers the decision in detail.

Does GlobalProtect work with Prisma Access and with on-prem firewalls?

Yes, the same app connects to either. With an on-prem firewall, GlobalProtect is a classic remote-access VPN. With Prisma Access, it becomes the on-ramp that backhauls internet traffic to a Palo Alto cloud location for SASE inspection. The second use is where latency and the reconnect behavior matter most for a distributed workforce.

Can GlobalProtect control personal ChatGPT while allowing the corporate tenant?

Not by itself. That control requires traffic to reach a Palo Alto decrypt proxy with AI Access Security, CASB-X and Enterprise DLP licensed on top. dope.security performs the allow-corporate, block-personal tenant restriction on the device through Cloud Application Control and inspects the prompt content with Dopamine DLP, without an add-on. The walkthrough is in how to block personal ChatGPT.

How do I deploy an alternative alongside GlobalProtect without a rip-and-replace?

Push the dope.endpoint agent through the MDM you already use (Intune, Jamf or similar), leave GlobalProtect in place for private applications, and let dope.SWG take web, SaaS and AI traffic. Outreach Health reached 99 percent device coverage in one week this way, and a Fortune 100 customer scaled from 900 to over 18,000 devices at roughly 3,000 per week using a silent Intune deployment.