← Streaming
D

YouTube

Serious concerns
Google · 🇺🇸 United States
PolicyApp PermissionsNetwork TrafficFirmwareRegulatory
Technical details
Manufacturer: Google LLC

⚠️ The bottom line

In 2019 YouTube paid $170 million for tracking children. In 2025, a federal judge said Google knowingly "engaged in highly offensive conduct" collecting kids' data — and approved another $30 million settlement. Disney got fined $10 million for mislabeling kids' videos on YouTube, which let Google's trackers follow children anyway. Six years and $210 million in penalties later, the same fundamental violation keeps happening. The children whose data was collected in 2019 are now teenagers — and their younger siblings are being tracked too. A jury found YouTube liable for deliberately hooking a 6-year-old child. She developed anxiety, body dysmorphia, and suicidal thoughts. YouTube's vice president testified under oath that his own kids watch for hours every day and it's "good for them" — while the algorithm was found to have caused documented psychological harm to a child.

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
D
Parent company: Google (Alphabet)
PRISM participant since 2009, CLOUD Act jurisdiction (US), $170M COPPA violation (YouTube Kids) +1 more
4 structural risks · 37 products →
Spying
4/4 EXTREME
Is someone spying on me?
Kids at risk
Data Sharing
3/4 HIGH
Who gets my data?
Security
2/4 MODERATE
Is it actually secure?
Honesty
4/4 EXTREME
Can I trust what they say?
Kids at risk
REPLACE Extreme risk. Look for alternatives or lock down hard.
13Contradictions
2Critical
6High
5Medium
17Sources
Findings by concern
Spying 4/4 EXTREME 7 findings
⚠️ criticalmarketing vs regulatory
A jury found YouTube liable for deliberately hooking a 6-year-old child. She developed anxiety, body dysmorphia, and suicidal thoughts. YouTube's vice president testified under oath that his own kids watch for hours every day and it's "good for them" — while the algorithm was found to have caused documented psychological harm to a child.

What they claim: YouTube claims its recommendation system "helps users discover content they'll enjoy and actively reduces harmful content"

What we found: On March 25, 2026, a Los Angeles jury found YouTube liable in a first-of-its-kind social media addiction case. The plaintiff started using YouTube at age 6 and developed anxiety, body dysmorphia, and suicidal thoughts. YouTube was assigned 30% responsibility. Jurors awarded $3M compensatory + $900K punitive. YouTube VP Cristos Goodrow testified his own kids use it for hours daily and it's "good for them."

⚡ highpolicy claims vs app permissions
Alexander Hanff, a privacy researcher, filed a formal complaint that YouTube's ad-blocker detection qualifies as spyware under EU law. The Irish Data Protection Commission agreed and opened a formal probe in 2025. YouTube's scripts reach into your browser to detect what extensions you've installed — the exact behavior the ePrivacy Directive was written to prevent. If you refuse to let YouTube scan your browser, YouTube blocks you from watching videos entirely. The choice is: let us surveil your browser, or leave.

What they claim: YouTube's privacy policy states it respects user choices about data collection and operates transparently.

What we found: In 2023-2024, YouTube deployed anti-ad-blocker technology that privacy researcher Alexander Hanff formally complained violated Article 5.3 of the EU ePrivacy Directive — the same provision governing spyware. The Ireland DPC accepted the complaint in 2025 and opened an investigation. YouTube's detection scripts probe browsers to detect ad-blocking extensions without consent.

⚡ highpolicy claims vs app permissions
Caleb Cain told the New York Times that YouTube's autoplay recommendations gradually pulled him from self-help videos into white supremacist content over several years. A PNAS study tested 100,000 fake accounts and confirmed: the deeper you follow recommendations, the more extreme they get. YouTube's own leaked internal research showed the company knew but prioritized watch time. In 2026, researchers in Turkiye found the algorithm could radicalize young conservatives "in a very short time." YouTube calls it a "discovery engine." Researchers call it a rabbit hole.

What they claim: YouTube states its recommendation system helps users discover content they'll enjoy and actively reduces harmful content.

What we found: A 2023 PNAS audit using 100,000 sock puppet accounts found YouTube's algorithm recommends progressively more extreme content. Caleb Cain told the New York Times he was pulled into white supremacist ideology through autoplay recommendations. YouTube's own leaked internal research showed the company knew but prioritized engagement metrics. A 2026 Turkiye study found the algorithm could radicalize conservative young people in a very short time.

⚡ highpolicy claims vs network analysis
In 2022, 40 US state attorneys general fined Google $391.5 million for tracking people's locations even after they explicitly turned off Location History. Arizona's case revealed internal emails where employees called their own practices "misleading." One Google engineer wrote that the only way to truly stop location tracking was to put your phone in a Faraday cage. YouTube uses this data to target ads at the zip code level. When you turn off location tracking, Google doesn't stop — it just uses Wi-Fi, IP addresses, and cell tower data instead. The off switch doesn't turn anything off.

What they claim: Google says YouTube collects general area location data and users can control location sharing through device settings.

What we found: Google was fined $391.5 million by 40 US state attorneys general in 2022 for tracking users' locations even after they turned off location history. Arizona's case revealed internal emails where employees called their own practices misleading. A Google engineer wrote the only way to stop tracking was to hide your phone in a Faraday cage. YouTube uses location data for zip code level ad targeting.

⚡ highpolicy claims vs app permissions
While Congress debated banning TikTok as a surveillance threat, Google positioned YouTube Shorts as the safe American alternative. But Shorts collects the same behavioral data — scroll speed, dwell time in milliseconds, swipe patterns, device fingerprints. The difference: Google connects your Shorts viewing to your Gmail, Google Maps location history, Chrome browsing, Android phone data, and entire search history. TikTok is a standalone surveillance app. YouTube Shorts is a surveillance app connected to everything else Google already knows about you. Congress worried about the wrong app.

What they claim: Google positioned YouTube Shorts as a privacy-respecting alternative to TikTok during the ban debates.

What we found: YouTube Shorts collects the same behavioral data as TikTok: scroll timing, dwell time in milliseconds, swipe patterns, audio fingerprints, device identifiers. But Google's advertising ID links Shorts behavior to Gmail, Google Maps, Chrome browsing, and Android device data — creating a cross-platform profile far more comprehensive than TikTok's standalone surveillance.

⚡ highmarketing claim vs app permissions
To label AI videos, YouTube has to analyse every video ever uploaded at a forensic level — scanning pixels for watermarks, running detection models, and matching faces against a database of every adult who requests it. Creators uploading a cooking video are now having their content run through the same analysis pipeline. YouTube says the labels don't affect anything, but research found that visible AI labels discourage viewers from engaging — so the label is a punishment even if the algorithm doesn't change. YouTube built a mass video surveillance system and called it "transparency."

What they claim: YouTube says the AI auto-labelling system uses "internal signals" to detect AI-generated content, with labels having no impact on recommendations or monetisation

What we found: To detect AI content, YouTube must analyse every video's visual characteristics at a deep level — forensic-grade AI analysis on all uploaded content. The platform scans for C2PA metadata and SynthID watermarks, plus undisclosed proprietary detection methods. YouTube simultaneously expanded deepfake face-matching to all adults, scanning all video content against face databases. Research shows visible AI labels discourage viewer engagement regardless of algorithmic treatment.

⚫ mediumpolicy vs app
YouTube stopped trusting creators to label their own AI videos. Now it scans every upload automatically — checking whether you used AI, even if you say you didn't. Google's models inspect your content, frame by frame, before it goes live. They call it "improving AI labels." It's really a system where Google decides what's real and what isn't — for every video on the platform.

What they claim: YouTube states creators are responsible for disclosing AI-generated content in their uploads

What we found: YouTube is now auto-scanning all uploaded videos to detect AI-generated content — regardless of whether the creator discloses it. Every upload is analysed by Google's detection models. This is content surveillance at scale: Google inspecting what you made and how you made it, applied to every video on the platform.

Data Sharing 3/4 HIGH 3 findings
⚫ mediumpolicy claims vs policy claims
YouTube Premium costs $13.99 a month and promises an "ad-free experience." What it removes are the visible ads. Google still collects every video you watch, every search, your location, your device fingerprint — identically to free users. That data feeds Google's $307 billion advertising business across every other Google service. You pay $167.88 per year to hide the ads from yourself while Google continues selling the profile those ads would have been based on. Premium buys a cosmetic improvement, not a privacy one.

What they claim: YouTube Premium is marketed as an ad-free experience implying privacy benefits.

What we found: YouTube Premium removes visible ads but does not reduce data collection. Watch history, search queries, device information, and location data are still collected identically to free users. Google's 2024 10-K confirms Premium revenue is separate from advertising, meaning Google double-dips: subscription fee plus behavioral data powering its $307 billion ad business.

⚫ mediumpolicy claims vs app permissions
YouTube requires creators to hand over government ID, tax details, and submit every video to Content ID fingerprinting. In return, YouTube reserves the right to run ads on any video — even channels that chose not to monetize. YouTube profits from your content whether you want it to or not. During periodic "purges," YouTube has terminated thousands of creator accounts overnight, erasing years of videos and subscriber relationships without meaningful explanation or appeal. Creators build their livelihoods on a platform that fingerprints their work, monetizes it without permission, and can erase everything with no recourse.

What they claim: YouTube says it supports creators and provides tools for them to manage their content and privacy.

What we found: YouTube requires government ID and tax details from creators, submits all videos to Content ID fingerprinting, and reserves the right to run ads on any video — even channels that chose not to monetize. During periodic purges, YouTube terminated thousands of accounts overnight, erasing years of content without meaningful explanation or appeal.

⚫ mediummarketing claim vs third party research
YouTube says it's labelling AI content to protect you from deception. Kapwing found that 278 channels uploading nothing but AI slop have racked up 63 billion views and an estimated $117 million in ad revenue — money YouTube shares in. When researchers created a fresh account, 21% of the first 500 recommended videos were AI-generated. One channel in India earns an estimated $4.25 million a year from pure AI slop. YouTube's algorithm is the engine that recommends this content to billions of people. They're labelling the flood while profiting from it.

What they claim: YouTube frames AI auto-labelling as a transparency measure to combat misleading content and protect viewers from AI-generated deception

What we found: Kapwing found 278 channels uploading nothing but AI slop, collectively drawing 63 billion views, 221 million subscribers, and an estimated $117 million in annual ad revenue — revenue YouTube takes a cut of. A simulation of a new user's feed found 21% of the first 500 videos were AI-generated and 33% were brainrot. YouTube's algorithm actively recommends this content to new users.

Honesty 4/4 EXTREME 3 findings
⚠️ criticalpolicy claims vs regulatory findings
In 2019 YouTube paid $170 million for tracking children. In 2025, a federal judge said Google knowingly "engaged in highly offensive conduct" collecting kids' data — and approved another $30 million settlement. Disney got fined $10 million for mislabeling kids' videos on YouTube, which let Google's trackers follow children anyway. Six years and $210 million in penalties later, the same fundamental violation keeps happening. The children whose data was collected in 2019 are now teenagers — and their younger siblings are being tracked too.

What they claim: YouTube's COPPA compliance page states it does not knowingly collect personal information from children under 13 without parental consent.

What we found: In 2019, the FTC fined Google $170 million for collecting children's data on YouTube for targeted ads. In 2025, a judge ruled Google knowingly engaged in highly offensive conduct, leading to a $30 million class action settlement. Disney was fined $10 million in December 2025 for mislabeling children's content which let Google's trackers collect data on kids.

⚡ highmarketing vs third party research
40% of videos YouTube recommends after a toddler watches Cocomelon are AI-generated slop. The top AI content farms make over $4 million a year from it. 230 child development experts told YouTube to stop. YouTube's response: invest $1 million in an AI studio that makes videos specifically for babies.

What they claim: YouTube claims to have "high standards for the content in YouTube Kids"

What we found: On April 1, 2026, 230+ organisations and experts including Jonathan Haidt demanded YouTube stop recommending AI-generated content to children. NYT investigation found 40% of videos recommended after watching preschool content were AI-generated. Top AI slop channels earned $4.25M+ annually. YouTube invested $1M in Animaj, an AI studio creating AI videos targeting babies and toddlers.

⚫ mediummarketing claim vs app permissions
YouTube now auto-labels creators' AI videos with a prominent badge — even overriding their objections. At the same time, Google just unveiled Gemini Omni and Veo to generate hyper-realistic video, and integrated them directly into YouTube Shorts via Dream Screen. Google is building the AI video factory and policing the AI video label. The recommendation algorithm that decides what 2 billion people watch every day is itself powered by AI, but that never gets a label. Creators bear the transparency burden. The platform that profits from AI demands everyone else be honest about using it.

What they claim: YouTube says it is implementing automatic AI content detection and labelling to "improve transparency" for viewers

What we found: YouTube's own recommendation algorithm — which decides what 2 billion users watch — is itself an AI system, never labelled as AI-mediated. Google aggressively builds AI video generation tools (Veo, Dream Screen, Gemini Omni) integrated directly into YouTube, while requiring creators to bear the transparency burden. Videos made with YouTube's own tools get permanent, unappealable labels, while third-party AI creators can dispute theirs.

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