← Operating System
F

Android / Google Play Services

Google knows where you are, even when you tell it not to. It settled for $391 million over that. Then kept doing it.
Fail
Google · 🇺🇸 United States
PolicyApp PermissionsNetwork TrafficFirmwareRegulatory
Technical details
Manufacturer: Google

⚠️ The bottom line

Professor Douglas Leith at Trinity College Dublin found your Android phone sends device identifiers and telemetry to Google every 4 minutes — even after you turn off all data sharing options. An idle Android sends 20 times more data to Google than an idle iPhone sends to Apple. Google collects your IMEI, hardware serial number, SIM serial, phone number, and advertising ID. You cannot stop it. There is no toggle. The data flows to Google every 240 seconds whether you consent or not. A billion Android users have no opt-out from continuous surveillance by the company that makes their phone's operating system. Google offers you a "Reset advertising ID" button — implying you can break free from tracking. But Google also collects your permanent hardware serial number, IMEI, and SIM serial number with every telemetry transmission. Resetting your advertising ID is like changing your hat while wearing a name tag. The new ID is immediately re-linked to your permanent identifiers. Google built the privacy control. Google also built the system that makes the privacy control meaningless. The reset button exists so Google can say you have a choice. The permanent identifiers exist so the choice doesn't matter.

Legal jurisdiction
🇺🇸 United States (headquarters)
CLOUD Act read more →
US govt can demand your data from this company even if stored overseas
FISA §702 / PRISM read more →
NSA collects stored emails, photos, messages without individual warrants
Geofence warrants read more →
Police can demand location data for everyone near a crime scene
Spying
4/4 EXTREME
Is someone spying on me?
Data Sharing
4/4 EXTREME
Who gets my data?
Security
4/4 EXTREME
Is it actually secure?
Honesty
4/4 EXTREME
Can I trust what they say?
REPLACE Extreme risk. Look for alternatives or lock down hard.
Use Linux Mint instead
Zero telemetry, rejected Snap, community-funded
See report →
28Contradictions
12Critical
15High
1Medium
25Sources
Findings by concern
Spying 4/4 EXTREME 9 findings
⚠️ criticalpolicy claims vs network analysis
Professor Douglas Leith at Trinity College Dublin found your Android phone sends device identifiers and telemetry to Google every 4 minutes — even after you turn off all data sharing options. An idle Android sends 20 times more data to Google than an idle iPhone sends to Apple. Google collects your IMEI, hardware serial number, SIM serial, phone number, and advertising ID. You cannot stop it. There is no toggle. The data flows to Google every 240 seconds whether you consent or not. A billion Android users have no opt-out from continuous surveillance by the company that makes their phone's operating system.

What they claim: Google Privacy Policy: "You're trusting us with your information. We work hard to protect your information and put you in control." Android setup presents "Send usage and diagnostic data" as a user choice.

What we found: Prof. Douglas Leith (TCD, Mar 2021): even with Usage & Diagnostics off, Pixel sends ~1.2MB telemetry to play.googleapis.com/log/batch at startup. android.googleapis.com/checkin transmits IMEI, hardware serial, SIM IMSI, WiFi MAC, AndroidId, Droidguard key — regardless of settings. Data sent every 255 seconds (4.25 min) on average when idle. Google's own toggle text admits: "Turning off this feature doesn't affect your device's ability to send information needed for essential services."

⚠️ criticalpolicy claims vs regulatory findings
Google built Sensorvault — a database tracking the location of 592 million people's devices every two minutes. Police used "geofence warrants" to demand the identity of everyone near a crime scene. In 2020, Google received 11,554 geofence warrants. Zachary McCoy, a Florida man, was falsely suspected of burglary because his Google location data showed he biked past the burglarized house. He faced a potential felony investigation. An innocent person's daily bike ride, recorded by Google, nearly put them in prison. Google eventually limited Sensorvault access in 2024 — after years of handing over innocent people's data to police.

What they claim: Google positions location data collection as improving user experience. States users can control their data.

What we found: Google maintained "Sensorvault" database with detailed location records from 592 million accounts. Location logged every two minutes from Android devices. Police used geofence warrants: 982 in 2018, 8,396 in 2019, 11,554 in 2020. By 2021, 25% of all law enforcement requests to Google were geofence. Jorge Molina wrongfully arrested for murder in Arizona, jailed one week after Sensorvault placed his phone near crime scene. US Fifth Circuit ruled geofence warrants "categorically prohibited by the Fourth Amendment."

⚠️ criticalpolicy claims vs regulatory findings
Texas Attorney General Ken Paxton proved that Google's location tracking continued even when users turned it off, that Incognito mode was deceptive, and that facial recognition data was collected without consent. In 2024, Google agreed to pay Texas $1.375 billion — the largest privacy settlement by a single state in US history. The state proved what researchers had been saying for years: Google's privacy settings are theater.

What they claim: Google represented that Location History controls, Incognito mode, and biometric data handling provided meaningful privacy protections.

What we found: May 9, 2025: Texas AG Ken Paxton secured $1.375B settlement — largest single-state privacy settlement against Google ever. Alleged: Google collected geolocation via Maps even with Location History disabled; Chrome Incognito still tracked activity; Google Photos and Assistant captured voiceprints and facial geometry without consent violating Texas biometric law. Google did not admit liability.

⚡ highpolicy claims vs network analysis
Google says it only collects basic info to make your phone work, but an idle Android phone sends 20 times more data to Google than an iPhone sends to Apple.

What they claim: Google: "Collection of limited basic information, such as an IP address, is necessary to deliver content." Frames Android data collection as necessary for core functionality.

What we found: Leith (2021): During first 10 min of startup, Pixel sends ~3.6MB to Google (vs 42KB iPhone to Apple — nearly 100x more). Idle: ~1MB every 12 hours (vs 52KB iPhone). Scaled to 129M US Android users: ~1.3TB of handset data every 12 hours from idle phones alone. Pre-installed YouTube, Chrome, Docs, SafetyHub, Clock, Search bar all phone home without being opened.

⚡ highpolicy claims vs app permissions
Google says it never sells your personal data. Technically true — Google doesn't sell the data directly. Instead, it uses everything your Android phone collects (location, app usage, searches, browsing, contacts) to power a $224 billion annual advertising business. Advertisers pay Google to target you with precision. Whether that counts as "selling your data" depends on whether you think selling access to your attention profile is meaningfully different.

What they claim: Google: "We never sell personal information" and gives users "transparency and control over their ad experiences."

What we found: Schmidt/Vanderbilt (2018): ad-related data comprised 46% of all requests to Google from Android. Idle Android with Chrome sent nearly 50x as many data requests/hour as iOS with Safari. Schmidt documented Google can identify specific users by combining "user-anonymous" advertiser data with collected data. Google monetizes through 00B+ annual ad business. As bipartisan senators noted, hundreds of firms in real-time bidding receive sensitive info including device IDs, cookies, location, demographics.

⚡ highpolicy claims vs network analysis
Every major Android phone manufacturer collects your installed app list and usage data with no opt-out — the only Android variant that doesn't spy on you removes Google entirely.

What they claim: Google markets Android as open platform with user choice and freedom.

What we found: Liu, Patras, Leith (PLOS ONE 2023): Samsung, Xiaomi, Huawei, Realme, LineageOS all transmit substantial data to Google and third parties even idle. All collect full list of installed apps (revealing mental health, religion, dating interests). Xiaomi sends details of all app screens viewed including call timing to Singapore servers. Samsung has pre-installed Microsoft/LinkedIn telemetry. Huawei connects to avast.com and 360safe.com. No opt-out from any of it. Only /e/OS (strips all Google) sent zero telemetry.

⚡ highpolicy claims vs firmware analysis
When you turn Wi-Fi off on an Android phone, a default setting called "Wi-Fi scanning" continues scanning for nearby networks in the background — to improve location accuracy. Google uses nearby Wi-Fi access points to triangulate your position even when you've explicitly turned Wi-Fi off. The setting is buried under Location > Scanning. Turning off the radio doesn't turn off the surveillance.

What they claim: Users expect turning off Wi-Fi stops Wi-Fi-based location tracking.

What we found: Android's "Wi-Fi scanning" (Settings > Location > Location Services > Wi-Fi Scanning) scans for networks even when Wi-Fi radio turned off. Enables location via Google's wardriving database. Enabled by default, buried in sub-menu most users never find. Google forced to create opt-out for network owners ("_nomap" SSID suffix) in 2011 after EU concerns.

⚡ highpolicy claims vs network analysis
Google Maps Incognito mode creates the impression of privacy. But a Google engineer admitted in the Texas AG lawsuit depositions that Incognito mode in Maps still transmits your location to Google — it just stores it under a randomised identifier instead of your account name. The $1.375 billion Texas settlement specifically cited this as deceptive. "Incognito" meant Google knew where you were but stored it under a pseudonym instead of your name.

What they claim: Google Maps Incognito: "When Incognito mode is on, Maps won't save your browsing or search history to your account or send notifications."

What we found: Incognito doesn't disable device-level location services. GPS, Wi-Fi, Bluetooth data continue flowing to Play Services and Maps. Norwegian Consumer Council (2023): 94% of mapping services transmit location metadata regardless of privacy mode. Google admits Incognito "does not affect how your activity is used by internet providers, other apps, voice search, and other Google services." Location from Incognito stored with session identifier instead of account name — still collected and stored.

⚡ highpolicy claims vs app permissions
Google's "privacy improvement" to move location to your phone also deleted years of data and conveniently made the database police were using for warrants harder to access.

What they claim: Google (Dec 2023): Location History moving to on-device storage, framed as giving users "more control over their data."

What we found: Migration (2024-2025) auto-deletes all location history older than 90 days if users fail to act before arbitrary deadline. Default auto-delete quietly shortened from 18 months to 3 months. Timeline no longer viewable on web — only single phone. Some users reported data deleted despite choosing to keep it. The Register noted timing coincided with mounting legal pressure over Sensorvault geofence warrants — conveniently making police database harder to access.

Data Sharing 4/4 EXTREME 6 findings
⚠️ criticalpolicy claims vs network analysis
Google offers you a "Reset advertising ID" button — implying you can break free from tracking. But Google also collects your permanent hardware serial number, IMEI, and SIM serial number with every telemetry transmission. Resetting your advertising ID is like changing your hat while wearing a name tag. The new ID is immediately re-linked to your permanent identifiers. Google built the privacy control. Google also built the system that makes the privacy control meaningless. The reset button exists so Google can say you have a choice. The permanent identifiers exist so the choice doesn't matter.

What they claim: Google developer guidelines: Android Advertising ID (GAID) is "user-resettable" and apps must not "bridge Advertising ID resets." Marketed as privacy protection.

What we found: Leith (2021): android.googleapis.com/checkin simultaneously transmits resettable RDID/Ad ID alongside hardware serial, IMEI, SIM IMSI, AndroidId (persistent, requires factory reset), WiFi MAC, Droidguard key. Re-linking a reset ad ID is trivial. Liu, Patras, Leith (PLOS ONE 2023) confirmed: "Samsung, Xiaomi, Realme and Google all collect hardware identifiers as well as resettable identifiers" — "This largely undermines the use of user-resettable advertising identifiers." Only 2.08% of users opt out globally (Singular).

⚠️ criticalpolicy claims vs network analysis
In 2017, Quartz journalist Keith Collins discovered that even with location services turned completely off, no SIM card inserted, and after a factory reset, Google was secretly collecting cell tower identifiers from Android phones and transmitting them to Google's servers. The phones were tracking their location with every setting turned off and every obvious data path removed. Google confirmed the practice and said it would stop — after being caught.

What they claim: Google Android location settings implied that with location services disabled, the device would not transmit location data.

What we found: Nov 2017 (Quartz): since early 2017, all modern Android phones collected nearby cell tower addresses and sent to Google — even with location services completely disabled, no apps in use, no SIM card inserted. Even factory-reset devices transmitted when connected to Wi-Fi. Mechanism: change to Firebase Cloud Messaging. Google claimed data "never used or stored" and was "immediately discarded" — only stopped after Quartz contacted them. Users had no way to opt out.

⚠️ criticalpolicy claims vs regulatory findings
California Attorney General Rob Bonta proved that Google continued tracking users' locations even after they turned off "Location History" — because a separate, less visible setting called "Web & App Activity" kept recording it. Google used that hidden location data to target ads. In 2023, Google settled with California for $93 million. The setting that said "off" was a decoy. The real tracking happened somewhere else entirely.

What they claim: Google told users disabling Location History would prevent storage of location data.

What we found: Sep 14, 2023: California AG Rob Bonta announced $93M settlement. Google falsely told users turning off Location History stopped storage, but continued collecting through other sources. Also deceived users about opting out of location-targeted advertising. Google's fifth settlement over geolocation. Required auto-delete location data within 30 days and disclose location data used for ad personalization.

⚠️ criticalpolicy claims vs app permissions
The privacy labels in Google Play Store — the ones you check before installing an app — are written by app developers themselves. Google doesn't verify them. Researchers at Mozilla studied 40 popular apps and found 80% had significant discrepancies between their Play Store labels and their actual privacy policies. The labels told you one thing; the apps did another. Google built a trust system based entirely on self-reporting by the people who benefit from lying.

What they claim: Google Play Data Safety section tells users what data apps collect. Google: apps "are responsible for making complete and accurate declarations."

What we found: Mozilla Foundation (Feb 2023, "See No Evil"): nearly 80% of top apps had false or misleading Data Safety labels vs actual privacy policies. 40% of 40 top apps received "Poor" grade (Minecraft, Twitter, Facebook). TikTok and Twitter claimed they don't share with third parties — policies explicitly state they share with advertisers. Google exempts "service providers" and "anonymized" data. Google performs zero verification. None of top 20 apps correctly disclosed collection (2026 arxiv study). In 2024, Google unpublished 1.3 million apps for data issues yet labels remain self-reported.

⚠️ criticalpolicy claims vs network analysis
Google's own chief marketing officer, Lorraine Twohill, wrote in an internal email that Incognito mode "is not truly private." Google kept marketing it with a spy icon and language implying secrecy. A class action lawsuit resulted in a $5 billion settlement — then $5.4 billion after appeal. Internal documents showed Google employees knew the branding was misleading. Engineers proposed fixes. Management declined. For years, hundreds of millions of people opened Incognito tabs believing they were private, while Google's own executives wrote emails acknowledging they weren't.

What they claim: Chrome Incognito: "You've gone incognito" and "Now you can browse privately." Internal Google email from marketing chief Lorraine Twohill to CEO Pichai (2019): "limited in how strongly we can market incognito because it's not truly private."

What we found: Class action (Jun 2020) alleged Google tracked browsing in Incognito via Google Analytics, DoubleClick, ad technologies. Settled 2024 — Google agreed to delete billions of browsing records. No monetary payment to class members. Google's own marketing chief admitted internally it's "not truly private." Updated disclaimer now reads: "This won't change how data is collected by websites you visit, including Google." Over 1,000 individual damages claims filed in California.

⚫ mediumpolicy claims vs app permissions
Three out of four Android apps secretly contain Google tracking code that monitors how you use them — even if you said no to data collection.

What they claim: Google Play Data Safety requires apps to disclose data collection. Google states core Play services SDKs "do not collect any end-user data."

What we found: Exodus Privacy / Yale Privacy Lab: 75% of Android apps contain at least one third-party tracker; Google Firebase Analytics in 59% (154,355 apps analyzed). Firebase SDK auto-collects app-instance ID, advertising ID, uses cookie-like technologies. 2023 ACM study: almost half of apps continued engaging trackers after users declined consent. Data flows to app-measurement.com linked to device RDID, enabling cross-app profiling. Invisible to users.

Security 4/4 EXTREME 4 findings
⚠️ criticalprivacy policy vs app permissions
Google's AI can now log into your bank, your email, your insurance portal — using your saved passwords, without you touching the keyboard. It fills forms, clicks buttons, and completes transactions while you watch. Google says human reviewers can see these conversations. Your AI assistant just became an authorised user on every account you've ever saved a password for.

What they claim: Google says Gemini Intelligence features are "opt-in" with user controls, and the security framework ensures "your data stays private"

What we found: Auto Browse in Chrome on Android can access Google Password Manager and log into third-party websites on your behalf — the first time Google's AI has been granted delegated credential use on external sites. Once enabled, Gemini navigates sites, fills forms, and completes bookings using your saved passwords. Google's own Privacy Hub confirms human reviewers can access AI conversations, and the Gemini Apps Activity policy states data may be retained up to 18 months.

⚠️ criticalmarketing vs third party research
A security researcher hid instructions inside a Google Doc. When anyone asked Gemini to summarise it, the AI silently copied out years of email, calendar entries, and documents — and sent them to the attacker. No clicks. No warnings. No alerts. A separate team did the same thing through a calendar invite. You don't even have to open it. Gemini's "enterprise-grade security" couldn't stop a booby-trapped document from raiding your entire digital life.

What they claim: Google positions Gemini as safe for enterprise use with "layered defense strategy" against prompt injection and built-in security controls

What we found: Security firm Noma Labs demonstrated "GeminiJack" — a single poisoned Google Doc containing hidden instructions that causes Gemini to exfiltrate years of email, complete calendar histories, and entire document repositories with zero clicks, zero warnings, and zero DLP alerts. Separately, Miggo demonstrated prompt injection via Google Calendar invites that bypasses Gemini's privacy controls to exfiltrate private meeting data — zero interaction required from the target user.

⚡ highpolicy claims vs app permissions
Google Play Protect scans every app on your phone for malware — and sends your complete app list to Google's servers. Researcher Douglas Leith at Trinity College Dublin found that even with Play Protect disabled, the scan data is still transmitted. Google's security feature doubles as an inventory of every application you use. Disabling it in Settings doesn't stop the transmission — it just hides the interface.

What they claim: Google Play Protect marketed as keeping "your apps safe & your data private." "Scam detection on-device ensures user privacy while enhancing security."

What we found: Play Protect scans all apps on device including sideloaded — 350 billion apps/day globally. Google "may receive information about OS and apps installed" even for Play Store apps with some protections disabled. With "Improve harmful app detection" enabled, entire app binaries uploaded to Google. On by default. Even with scanning disabled, Google states it "may continue to receive information about apps installed through Google Play." Installed app list reveals sensitive interests (health, dating, religion, politics).

⚡ highpolicy claims vs firmware analysis
Google says your phone backup is encrypted, but Photos, Drive files, and some messages use keys Google holds — meaning Google and law enforcement can read them.

What they claim: Google markets Android auto-backup to Google Drive as secure. Implies data protected in cloud.

What we found: E2E encryption for backups only works on Android 9+ with lock screen set. Photos, Drive files, and MMS use Google-held keys — not E2E. Google can access and provide backup content to law enforcement with valid warrant for non-E2E categories. Similar to Apple's iCloud problem: backup containing message keys makes per-message encryption moot. Users not informed which categories are E2E and which aren't.

Honesty 4/4 EXTREME 9 findings
⚠️ criticalpolicy claims vs network analysis
Google told 2 billion Android users that turning off "Location History" would stop tracking their movements. It didn't. Google kept recording your location through a separate setting called "Web & App Activity" that most users didn't know existed. The Associated Press investigation (2018) proved it. Google settled for $391.5 million with 40 state attorneys general — at the time, the largest privacy settlement in US history. Arizona's AG called it "deceptive and dishonest." Google's response: rebrand the feature and bury the real off switch deeper in settings.

What they claim: Google support page: "You can turn off Location History at any time. With Location History off, the places you go are no longer stored."

What we found: Aug 2018 AP/Princeton: Google Maps created location snapshot whenever opened even with Location History off — separate "Web & App Activity" independently recorded location. Searches unrelated to location ("chocolate chip cookies") saved precise lat/long accurate to the square foot. Princeton researcher Gunes Acar demonstrated by carrying Android with History disabled; AP mapped his commute across New York. Google settled: Arizona $85M (Oct 2022), 40 states $391.5M (Nov 2022) — largest multistate privacy settlement in US history.

⚠️ criticalprivacy policy vs app permissions
Google quietly gave its AI permission to read your text messages, see your call history, and access WhatsApp — without asking. When people noticed, Google's own statements contradicted each other about whether you could turn it off. Users started reporting that Gemini was taking screenshots of their banking apps and health records every time they accidentally triggered it with the wrong button. The opt-out was buried in settings most people will never find.

What they claim: Google states users have meaningful control over Gemini's data access with ability to "turn off" features and manage permissions

What we found: On 7 July 2025, Gemini automatically activated access to Android Messages, Phone, WhatsApp, and system utilities by default — without clear opt-in consent. Google's own statements contradicted each other: one said the change applied "whether your Gemini Apps Activity is on or off," another said previously-disabled features would "remain off." Users reported "unauthorised screenshots" on support forums when Gemini was activated via voice or hardware button, capturing banking apps, health apps, and messaging threads.

⚡ highpolicy claims vs firmware analysis
Google Play Services is marketed as a security tool, but it's actually Google's main data pipeline — hoovering up data from calls, contacts, photos, email, and you cannot remove it.

What they claim: Google: Play Services "helps you keep your apps running securely, and quickly delivers new features." Presented as security and functionality benefit.

What we found: Leith: Play Services responsible for bulk of mandatory telemetry. android.googleapis.com/checkin links IMEI, serial, SIM IMSI, AndroidId, WiFi MAC, email, Droidguard key every 6 hours. play.googleapis.com/log/batch aggregates data from 50+ logging sources including CARRIER_SERVICES, DIALER, GOOGLE_NOW, DRIVE, PHOTOS, CALENDAR, CONTACTS, GMAIL, MESSAGING. Cannot be uninstalled on any certified Android device. Disabling breaks most phone functionality.

⚡ highpolicy claims vs network analysis
Professor Douglas Leith at Trinity College Dublin found that even when not actively using any Google app, a stock Android phone sends your location to Google 14 times per hour and transmits nearly 12 megabytes of data per day. A phone sitting on your desk, screen off, sends your location to Google every 4.3 minutes. In comparison, an idle iPhone sends location data 1.5 times per hour. Your Android phone is 9 times chattier than an iPhone at doing absolutely nothing.

What they claim: Google presents products as responsive tools that work when you use them. Implicit promise: when not using Google, Google isn't collecting.

What we found: Schmidt/Vanderbilt (2018): 11.6MB/day between Android device and Google. Dormant Android with Chrome in background sent location data 340 times in 24 hours (14x/hour). Schmidt: "A major part of Google's data collection occurs while a user is not directly engaged with any of its products." Ad data = 46% of all requests. Phone isn't waiting for you to use Google; Google uses the phone to watch you continuously.

⚡ highpolicy claims vs firmware analysis
Your Android phone comes with apps you can't delete that collect your data in the background — the only way to remove them requires developer tools most people have never heard of.

What they claim: Google promotes "Protected by Android" with Play Protect for security and "GPS permissions for control of personal information."

What we found: Android ships with Google apps (Play Services, Chrome, Gmail, Maps, YouTube) that cannot be uninstalled — only disabled. Play Services cannot be disabled without breaking phone. OEM bloatware runs in background collecting data when never opened. Many pre-installed apps have contacts, location, phone permissions by default. Xiaomi sends details of all app screens viewed including call timing to Singapore servers (TCD study). Huawei's Swiftkey sends usage to Microsoft. Removing bloatware requires ADB developer tools.

⚡ highpolicy claims vs firmware analysis
Google can silently install apps on your Android phone and has done so — Google Pay, Google Messages, and other apps have appeared on phones without user action. Google can also remotely delete apps: in 2010, Google removed two apps from users' phones after determining they were malicious, and in 2021 used Google Play Protect to disable apps. Your phone is not entirely your phone. Google retains a remote control that works without asking.

What they claim: Google markets Android as giving users control over their devices. Play Store focuses on user-initiated actions.

What we found: Google can silently install apps without user consent via Play Services. MassNotify incident (Jun 2021): Google remotely installed COVID-19 exposure notification app on Massachusetts phones without consent, without app icon, without notification. Google confirmed "intentional." Also has documented kill switch for remote app removal. Play Services operates at higher privilege than user apps — system-level component users cannot fully control.

⚡ highpolicy claims vs app permissions
Android is technically usable without a Google account, but you lose the app store, backups, and contacts — and Google is making it even harder by requiring its approval for all app installs.

What they claim: Google describes Android as open platform giving users choice and control.

What we found: Without Google account: no Play Store (primary app source), no automatic backup, no contact/calendar sync. YouTube loses personalization, Maps loses saved places. Sideloading requires technical knowledge. Starting 2026-2027, Google extending mandatory developer verification to ALL Android apps — even sideloaded apps will query Google verification server on certified devices. /e/OS and LineageOS are only real alternatives but run on limited hardware.

⚡ highpolicy claims vs app permissions
Android's permission system shows you some of the data apps can access, but researchers found 72% of dangerous permissions are hidden from users entirely.

What they claim: Android permission system presented as user control: apps must request dangerous permissions and users can grant/deny.

What we found: ECCWS study found 72% of dangerous permissions not shown to users in device settings. Apps can access data beyond declared permissions through Android system APIs. Pre-installed apps bypass normal permission flow with system-level privileges. Google Play Services itself holds virtually all dangerous permissions by default with no user ability to revoke.

⚡ highpolicy vs app
That notification you just read? Google rewrote it. Both Apple and Google are now using AI to summarise, reorder, and silently edit push notifications before they reach you. The app developer wrote one thing. You read another. Neither you nor the developer consented to the edit. Your phone is no longer a messenger — it's an editor with opinions.

What they claim: Google claims to deliver notifications as sent by app developers

What we found: Google is using on-device AI to summarise, reorder, and rewrite push notifications before users see them. Messages from apps are altered without the sender's or receiver's consent — Google decides what you read and in what order.

Latest Risks & Threats
New developments that compound existing privacy concerns. 1 emerging risk.
RISK Gemini Intelligence — OS-level AI with credential delegation and screen capture ⚠️ Surveillance Announced 2026-05-12
Google announced "Gemini Intelligence" at The Android Show on 12 May 2026, arriving late June on Pixel 10 and Galaxy S26. Auto Browse navigates third-party websites using your saved passwords from Google Password Manager — the first time Google AI has delegated credential use on external sites. Screen Context captures whatever is visible when Gemini activates (banking apps, health records, messages). Rambler processes all voice dictation system-wide through Gemini, partially cloud-based. Connected Apps integration gives Gemini access to Gmail content, Calendar, call logs, installed apps, and device usage patterns. Digital Watch Observatory flagged "privacy concerns rising over Gemini's on-device data access." Android Authority editorial: "Gemini Intelligence is the future of Android, and I already hate it."
Sources
What happened to real people
Documented incidents involving Google products and user data.
Jorge Molina jailed 6 days for murder via geofence warrant based on Google Sensorvault location data. Lost job, car, reputation. Charges never filed. [source]
PRISM participant since 2009. NSA collects stored communications. FBI conducts warrantless 'backdoor searches' of American data using names and email addresses. [source]
Google received 180 geofence warrants per week by 2019. Each warrant searches tens of millions of accounts. Supreme Court hearing constitutionality (Chatrie v. United States). [source]
What your data is worth to governments
Google complied with 235,000 government data requests in H1 2024. That's +530% over 10 years. Google has been a confirmed PRISM participant since 2009. Under this programme, the NSA collects stored communications. The company is legally prohibited from telling you. Jurisdiction: US (CLOUD Act, FISA Section 702, Patriot Act).
Documented: Jorge Molina jailed 6 days for murder via geofence warrant based on Google Sensorvault location data. Lost job, car, reputation. Charges never filed.
Documented: PRISM participant since 2009. NSA collects stored communications. FBI conducts warrantless 'backdoor searches' of American data using names and email addresses.
What is PRISM? · What is the CLOUD Act? · Transparency report
Sources