Can you trust your router? GL.iNet, OpenWrt, and where it can go wrong

GL.iNet is the practical default for most readers, but "runs OpenWrt" doesn't make a router safe. This guide shows which layers matter, what you can check yourself, and where the honest limits of that recommendation sit.

Can you trust your router? GL.iNet, OpenWrt, and where it can go wrong

Can you trust your router? GL.iNet, OpenWrt, and where it can go wrong

GL.iNet is the practical default for most readers, but “runs OpenWrt” doesn’t make a router safe. This guide shows which layers matter, what you can check yourself, and where the honest limits of that recommendation sit.

Who this guide is for

For readers considering (or already using) a GL.iNet router who are wondering whether “runs OpenWrt” is the same thing as “safe.” Also relevant if you’ve come across a cheap OpenWrt-style travel router on AliExpress and aren’t sure how to judge it. This guide is about how to evaluate trust in a brand, not “which brand is bad.”

What this guide solves and doesn’t

This isn’t a final verdict on GL.iNet, and not a blacklist of brands. No router is risk-free, and nobody can prove hardware or firmware trust with absolute certainty. What this guide does do: show which questions actually matter, what you can check yourself, and why “Chinese brand” or “runs OpenWrt” are both too simple to rely on alone.

The practical default

For most readers, GL.iNet remains the reasonable choice in its price class — not because it’s risk-free, but because it’s the most transparent option that’s also usable and affordable. The rest of this guide explains why, and where the honest limits of that recommendation sit.

Why “it runs OpenWrt” isn’t a guarantee

“This router runs OpenWrt” often gets treated as the end of the conversation: open source, therefore safe. That’s a misconception. OpenWrt is the software layer — but a router has more layers than that:

  • The application layer (the management UI, cloud features, extra tooling) — this is where a vendor most easily adds something you don’t want.
  • The WiFi chip firmware — almost always closed source, regardless of which OS runs above it.
  • The hardware itself — what’s actually on the board, regardless of what software you put on it.

A router can be OpenWrt-based and still unsafe, and a router can run closed firmware and still be fine — it depends on what happens in each layer, not the label. This is also the exact conclusion of our earlier look at the US router-ban debate: flag and brand name are weak shortcuts, behavior and control tell you more.

GL.iNet as a case study: what argues for trust

GL.iNet is an established company: founded in 2010, headquartered in Shenzhen, with its own factory there since 2018 (according to its own About page), and sold through official channels with real support — not an anonymous storefront-only name. A few concrete signals that matter:

  • A public security-advisory program. GL.iNet publishes dated security advisories with CVE numbers and credits the researchers, and says it discloses vulnerabilities within 90 days. Examples: an unauthenticated remote code execution via SID bruteforce and missing input validation (CVE-2024-39225 and CVE-2024-39226, August 2024), command injection after login (CVE-2024-57391, April 2025), and firmware-verification flaws in its KVM products (April 2026). That sounds alarming, but it’s actually a good sign: issues get disclosed and patched rather than hidden. A brand with zero ever-reported CVEs is more often unaudited than perfectly secure.
  • Firmware based on OpenWrt, with their own layer on top. The OpenWrt base and the GPL parts are published as source, so it isn’t a fully closed fork. GL.iNet’s own applications and web interface are not: its published source release for the Beryl 7 says those proprietary parts are excluded.
  • Years of community scrutiny — but per model, not automatically for every new product. This isn’t a vague signal: SRLabs, a respected Berlin-based security research firm, builds and maintains blue-merle, a real anonymity-hardening package (IMEI changing, MAC and BSSID randomization, wiping MAC-address logs) specifically for GL.iNet’s older 4G Mudi (GL-E750). That’s serious, named independent scrutiny, not just hobbyist chatter. But the newest flagship — the Mudi 7 (GL-E5800), launched in January 2026 — hasn’t had that depth of attention yet: a GitHub issue asking SRLabs to extend blue-merle to the Mudi 7 is still open. In July 2026 SRLabs wrote there that it now has a Mudi 7 and has run first tests, but that the current blue-merle doesn’t support it out of the box (status 1 November 2026). Community trust is real for GL.iNet as a track record, but it builds up per model — a brand-new product simply hasn’t had that scrutiny yet.

The honest complication

Two things that shouldn’t be glossed over:

  • GL.iNet is a Chinese company, headquartered in Shenzhen (mainland China), with an office in Hong Kong and a US company. That puts it under China’s National Intelligence Law, which requires every organisation to “support, assist and cooperate with” national intelligence work (Article 7) and lets intelligence agencies ask for that help (Article 14). That’s not theoretical, and it applies to any Chinese hardware brand, not specifically to GL.iNet.
  • An old OpenWrt forum thread (2021) alleged that GL.iNet’s firmware (then based on OpenWrt 19.07) was leaking DNS requests to Chinese domains (chinauos.com/uniontech.com), even with DNS-over-TLS and a VPN active. A GL.iNet staff member replied in the thread and asked for the logs; the poster didn’t share them publicly. One user who ran a packet capture on their own GL.iNet router saw nothing suspicious, and the thread closed with nobody having reproduced the issue. That’s not a confirmed leak — but it was never fully settled either. Unresolved, not disproven.

Can you verify this yourself?

Yes, to a degree — through layered methods, none of which is conclusive on its own:

  1. Hash the WiFi firmware blob and compare it against the chip vendor’s own published copy — the sources OpenWrt itself builds from, such as the linux-firmware project, Qualcomm’s ath11k-firmware repository, or MediaTek’s files in the mt76 driver repository. If the hash matches exactly, you’re now trusting the chip vendor (Qualcomm/MediaTek) instead of GL.iNet — that narrows the trust surface, it doesn’t eliminate it.
  2. Monitor network traffic from a separate device on the same network segment, with every OpenWrt-side feature disabled. This catches active “phone home” behavior, not something waiting for a trigger.
  3. Physical teardown / PCB inspection, ideally against a published schematic or reference unit — hobbyists do post this for GL.iNet hardware (for example board photos of the Mudi 7 on the OpenWrt forum).
  4. Years of unpaid community scrutiny (see above) — not a technical test, but a real signal.

The honest epistemic point: you can never prove the negative. There’s no way to cryptographically prove a backdoor doesn’t exist — you can only stack independent layers of evidence until the residual doubt is small enough to act on. That fits this site’s core conclusion: no brand is flawless, some brands just handle flaws better.

Could a backdoor come through the WiFi firmware, even without GL.iNet’s own software?

Yes — and this is the less obvious, more serious risk. Research from Google Project Zero (“Over The Air,” 2017) showed that a Broadcom WiFi chip could write directly into the host’s RAM over PCIe, including the kernel’s code — bypassing the operating system entirely, because no IOMMU was in place to stop it. That was tested on two phones (Nexus 6P and Galaxy S7 Edge), not on routers, and whether a given router fences off its WiFi chip this way differs per chip. But it shows that a compromised WiFi firmware blob isn’t just “the radio is untrustworthy” — it can mean full device compromise, invisible to anything OpenWrt itself can see.

This is exactly why flashing vanilla OpenWrt doesn’t close this gap — the blob stays, regardless of what runs on the main chip.

Flashing vanilla OpenWrt: a real middle step, not a fix

Many GL.iNet models are supported by the official OpenWrt releases, so you can flash “vanilla” (unmodified) OpenWrt instead of GL.iNet’s firmware: OpenWrt 25.12.6 lists about 40 GL.iNet models, including the Beryl AX, Flint 2, Beryl 7 and the original Mudi. That removes GL.iNet’s own application layer — the layer most likely to carry a backdoor or unwanted telemetry — and replaces it with firmware built through the OpenWrt project’s own, publicly reviewable process.

This doesn’t hold automatically for every model. The Mudi 7 specifically — 5G plus WiFi 7, built on a Qualcomm Dragonwing platform according to GL.iNet — currently has no official upstream OpenWrt support (not in OpenWrt 25.12.6). GL.iNet’s support, quoted in an OpenWrt forum thread in May 2026, says stock OpenWrt support depends on chip makers releasing open-source drivers, and gives no date. Newer, more complex combo devices (5G + WiFi 7 in one chip family) are a genuinely different situation from a simpler, longer-established model — check per model, not per brand, whether vanilla OpenWrt is actually available.

What it doesn’t fix: the firmware that runs on the WiFi chip itself stays a closed binary. Vanilla OpenWrt uses open-source WiFi drivers on the main chip where they exist (for example mt76 for MediaTek and ath11k for Qualcomm), but those drivers still load a closed firmware blob onto the radio. On models with a 4G/5G modem, the modem firmware stays closed too.

Conclusion: a real, meaningful improvement — not a way to bring the risk to zero.

The red flags: unbranded AliExpress routers

The bigger practical risk isn’t GL.iNet — it’s the anonymous “OpenWrt-style” travel routers sold on AliExpress under shifting brand names. Concretely, watch for:

  • No findable company behind the name — no real website, no support channel, just an AliExpress storefront.
  • No security advisory ever published — not because the brand is flawless, but because nobody’s ever checked, or the brand reports nothing.
  • Unverifiable firmware claims — “based on OpenWrt” with no source, no build process, no way to check.
  • No accountable jurisdiction — if something goes wrong, there’s nobody to hold responsible.
  • Store rating isn’t a security signal — a 4.8-star AliExpress storefront tells you about shipping speed and customer service, nothing about firmware safety. A related warning: firmware-security firm Eclypsium wrote in 2026 that counterfeit routers and switches can run modified or never-updated firmware, in some cases with hardcoded credentials or undocumented remote access.

A real EU alternative already exists: Turris

If the question is “could a European company solve this by rebranding GL.iNet” — the answer is: rebranding alone is cosmetic unless the company also builds and controls the firmware itself. But the stronger model already exists, and goes further than rebranding: Turris, made by CZ.NIC (the association that has run the Czech Republic’s .CZ domain registry since 1998 — not an anonymous vendor, but an established organization).

  • Hardware made in the Czech Republic, according to Turris (“a secure router made in the EU”), not just designed in the EU and built in China.
  • Firmware (Turris OS, OpenWrt-based) with all source code public, on CZ.NIC’s own GitLab and mirrored on GitHub, plus automatic updates.
  • CZ.NIC writes publicly about vulnerabilities that touch Turris (for example a 2020 blog post on a package-manager CVE and how Turris’s updater limits it) — the same “real issues, discussed openly” signal as GL.iNet, now backed by EU jurisdiction and EU manufacturing.

The honest limitation: Turris only makes home/business routers (Omnia, Omnia NG, Omnia NG Wired, the modular MOX line) — no compact travel router, and pricing sits clearly above GL.iNet’s entry tier. For travel use, there currently isn’t a full EU-made equivalent in the same price class — a real, still-unfilled gap, not something to paper over.

The practical checklist

Applicable to any brand, including one neither of us has checked yet:

  • Is there a findable company with a track record behind the name?
  • Does the brand ever publish security advisories/CVEs, or is it silent?
  • Is the firmware (partly) open source and checkable?
  • Does the device support flashing independent firmware?
  • Is there years of independent community scrutiny for this specific model?
  • Where is it manufactured, and does that change the trust chain for your situation?

When the stronger setup is worth it

For most home users, GL.iNet with stock firmware, a strong password, and unnecessary cloud features disabled (see the network threat profile guide) is more than enough. Consider the step to vanilla OpenWrt or Turris if:

  • you have a specific, concrete reason to distrust a given vendor (e.g. a profession with a real geopolitical threat model)
  • you already have the technical skill and interest to flash and maintain your own firmware
  • Turris’s higher price and lack of a travel form factor aren’t a problem for your use case

For most readers, this is overkill — and that’s a perfectly valid conclusion to land on.

Next step

Want to go further?