Compliance
Three compliance surfaces
One clear position
A compliance review of AI Gateway has three surfaces: what the Module touches on the device, what platform review requires, and what routed traffic is used for on the network.
constant foundation
Three facts that hold everywhere
Read this once. The rest of the page assumes it
1
What it touches
Three connection-level attributes
The Module reads three connection-level attributes: IP address, connection type, and approximate region. Nothing what is routed is written to or stored on the device
Not reachable by design: Files · Contacts · Photos · Location services · Browsing history · Cookies · Keystrokes · Screen content · App usage · Microphone / camera.
2
How it starts and stops
Consent-gated at the Module layer
Opt-in is enforced at the Module layer: no consent, no activation, no traffic. A single opt-out in settings stops it immediately and persists across sessions.
Every platform entry below references this mechanism rather than restating it. Where a platform adds a requirement, the entry states only the addition.
The consent screen is built from our template and states, in plain language, what is contributed, what it is used for, and that it can be turned off at any time.
3
Where the traffic goes
Publicly available data only for vetted AI
Routed requests are read-only fetches of public web pages, performed for vetted, KYC-verified buyers, for legal and age-appropriate use cases only.
The network surface is bounded in section 04.
Device surface
Fact 1 is the whole story
Only three connection-level attributes and nothing else is accessible by design. Full details on network data, stored data, and what’s visible on-device are documented separately
At review
Every platform's rules, mapped
Single-gatekeeper stores
One body certifies
Apple
iOS · iPadOS App Store Review
App Review Guidelines + Privacy Manifests
iOS · iPadOS App Store Review
App Review Guidelines + Privacy Manifests
5.1.1 (v)
App Privacy labels
Microsoft
Windows Store certification
Store Policies + Unwanted Software criteria
Windows Store certification
Store Policies + Unwanted Software criteria
10.5
No single gatekeeper
Obligations come from your channel
There is no universal Linux review. The rules follow the distribution channel you choose.
Linux
Flathub · Snap · distro repos Per distribution channel
Submission, sandbox & licensing rules
Flathub · Snap · distro repos Per distribution channel
Submission, sandbox & licensing rules
There is no Apple- or Microsoft-style review on Linux. Obligations come from the channel: Flathub / Flatpak, the Snap Store, or distro repositories (Debian-style).
Not unwanted software
The cardinal sin across Linux is undisclosed background network activity. Built directly against it: opt-in at the Module layer, revocable immediately, with persistent state.
Engine, not a store
The real review is where you ship
Unity
Engine & Asset store Asset Store rules if packaged
Destination store governs if shipped
Engine & Asset store Asset Store rules if packaged
Destination store governs if shipped
Unity is an engine, not a store — the real store review is wherever your game ships. Unity itself adds one duty: a third-party-data disclosure.
List AI Gateway as a third party in your privacy policy and link its policy. This is the single most important Unity action item. The disclosure is small and accurate because the Module collects only connection-level data.
The network surface
What your users' capacity is used for
This is the surface the industry category has historically failed on, so it gets the same treatment as store review: who decides, what’s checked, and what happens on failure
Who buys
Contractual, never anonymous
Every buyer clears KYC before receiving anything as anonymous demand is not served
Since 2021, demand has been routed through an exclusive partnership with Oxylabs, whose supplier standards this network is contractually held to — documented consent, ongoing vetting, termination for deviation.
What's allowed through
Read-only, public pages
Requests are read-only HTTP fetches of publicly accessible pages and all use cases are restricted to legal, age-appropriate purposes
No logins. No form submission. No content injection. No writes.
What's watched
Blocked as it happens
Traffic is monitored in real time so that anomalous patterns are blocked as they occur, not investigated afterward
Independent penetration testing by Blaze Infosec runs continuously, reviewable under NDA.
Scope
Stated here, shown in full elsewhere
This section bounds the network surface so a review can finish here. The demand chain itself – who the counterparties are and how the network is sourced – is documented in full.
What happens on failure — in both directions
A buyer that violates use-case restrictions loses access
Contractual termination, not a warning cycle
A partner whose integration degrades loses routing
Consent quality, disclosure accuracy and opt-out accessibility are re-audited after material changes. Violations discovered post-launch can end the partnership. Every partner in the pool has cleared the same review you are reading about — which is what protects your users from someone else’s integration, and their users from yours
Where the obligation sits
Two lists,
not shared responsibility
We provide
- The Module, with opt-in enforced at its layer
- The consent screen template and its required elements
- Written approval of your consent screen before production
- Platform declaration text for each submission path
- KYC on every commercial counterparty
- Real-time monitoring and enforcement on the network
- Penetration test reports, reviewable under NDA
You implement
- Presenting the consent screen before activation
- Keeping opt-out reachable in settings, permanently
- Submitting the declaration text with your build
- Telling us when your integration materially changes
- Confirming whether your Android builds carry GMS certification
What a reviewer can verify
Signals that hold under scrutiny
The market has trust baggage. We carry the standards that say we don’t. Below: the network at scale, then the attestations behind it.
GDPR & CCPA by design
Lawful basis is consent, collected before activation and withdrawable at any time.
Explicit, revocable consent
The in-app toggle provides immediate opt-out, persistent across sessions.
Real-time anomaly blocking
Misuse is detected and blocked as it happens, not after the fact.
KYC on every buyer
Every demand-side buyer is verified before access to the network.
Independent pen testing
Continuous testing by Blaze Infosec, reviewable under NDA.
Read-only, public only
Each node performs a single, public, read-only HTTP GET. Nothing is written to the device.
A+ provider
Vetted to the standards an industry leader demands of its suppliers.
Pentested by Blaze
Penetration-tested by independent security specialists.
AMTSO
Testing practices aligned with the industry's anti-malware standards body.
EWDCI
Committed to the industry's standard for ethical, consent-based web data.
ANSI/TIA-942-B
Data center infrastructure engineered to audited reliability tiers.
NIST 800-53
Security controls aligned with U.S. federal-grade requirements.
ISO/IEC 27001
Information security managed to the global enterprise standard.
ISO 14001
Environmental responsibility built into how we operate.
ISO 50001
Energy use measured, managed, and continuously improved.
An honest caveat
A correctly integrated app following the legal integration guide is well-positioned. But no third party can guarantee an individual review outcome, and we won’t claim otherwise.
This page argues integration-level compliance with each platform’s published rules. Store and channel review is discretionary and case-by-case — bandwidth-sharing components attract scrutiny.
