this month a japanese manga artist woke up with no google account. no gmail, no drive, no photos, nothing. the reason is he had uploaded private files of an old comic he drew to google drive. google's AI scanned, the AI flagged, the AI denied his appeal. no human ever looked at his case. years of private drawings: gone. every third-party service tied to his google login gone too.
this guy isn't an isolated case. he's just the one that went viral this week.
in 2022 in the united states, a software engineer named mark took a photo of his toddler son's groin, the kid had an infection, the pediatrician asked for the picture over telehealth to confirm the diagnosis. the photo got auto-backed-up by google photos. 48 hours later his account was suspended for "harmful content, potentially illegal." google filed a report with NCMEC (the us national center for missing and exploited children).
san francisco police investigated, concluded no crime had been committed, gave mark the police report. mark appealed to google with the police report in hand. google refused. mark lost 10 years of gmail, his contacts, every photo of his son's childhood, and his google fi phone number.
without his number and email, he couldn't even sign into the dozens of third-party services that depended on them.
same week, in houston, a dad named cassio. identical story. pediatrician, photo, ban, appeal denied.
in 2024, AI-generated CSAM reports automatically filed to NCMEC went up +1,325 % vs 2023. that's not 13x more pedophiles. it's classifiers scanning more, flagging more, while the humans who used to review are replaced by other classifiers.
and that's only the visible part. google scans your drive documents, your mail, your photos, your chrome history, your voice on google assistant, your location. not "scans to protect you." scans, indexes, monetizes. you're not the customer. you're the product. you've known this for 15 years and you keep paying in data because it's free in dollars.
the real problem: you own nothing
when you upload a file to google drive, you give the file to google. it's in the TOS. google can read it, move it, delete it, use it to train their models, and ban you if the algorithm decides it's sketchy. you have no recourse.
there's no human to call. there's no court. there's an appeal form, read by another AI, that replies "google stands by its decision."
with 2 billion gmail accounts, google does not have the human bandwidth to process appeals one by one. so an AI flags, an AI denies the appeal, and that's it. your account is dead.
is even worse than that: if you use "sign in with google" on 30 sites (and you do, github, slack, figma, notion, canva, etc.), losing google = losing 30 services. your 2FA codes on google authenticator? dead. your google fi number? dead. your entire digital life depends on a single algorithmic judgment, made in 0.3 seconds by a classifier.
you should not "follow the rules better." you should stop putting all your eggs in the hands of someone who can break the basket without notice.
exit #1: chrome → brave
chrome is a spy. every url you type, every page you visit, every minute you spend: sent back to google. google spent two years trying to kill ublock origin extension that blocks ads with manifest v3. and succeeded.
brave is chromium without google. same engine, same performance, but:
- 101 million monthly active users as of september 2025. crossed the 100M line faster than chrome did at its launch
- blocks ads and trackers by default → pages load 3 to 6x faster
- native anti-fingerprinting: brave randomizes the values of the APIs used to identify you across sessions, and blocks or modifies the chatty ones
- scored 143/156 on privacytests[.]org in 2025, highest among mainstream browsers. firefox: 80. chrome: 36.
just install it, import your bookmarks, go on with your life. you don't need to "technically" understand a thing.
exit #2: gmail → proton mail
gmail reads your mail. they denied it for 10 years, then admitted it, then re-denied, then re-admitted under filtered terms ("only to improve experience", translation: train gemini, sell ads, or tag your account for suspicion). in practice: your mail is not private.
proton mail:
- based in switzerland, a jurisdiction that bans data transfers outside the EU without judicial process
- end-to-end encryption by default. proton cannot read your mail, even if they wanted to
- zero-knowledge encryption: they don't even store your decryption keys. only your password generates them client-side
- no scanning, no classification, no AI deciding your fate
- free up to 500 MB. paid plans start at €4/month. over 100 million accounts in 2025
migrating takes 5mn. you create a proton account, use their "easy switch" to import gmail + contacts + calendar, set up a forward rule on gmail for the next 6 months as a transition layer, and update your address on the 20 critical accounts (bank, providers, taxes).
exit #3: google docs → fileverse docs (dDocs)
google docs: you write, google reads. your drafts are indexed. your internal templates can vanish if the AI decides. your collaborators are tracked via the "google login required" gate.
fileverse docs: the most mature decentralized alternative:
- end-to-end encryption by default. fileverse cannot read your documents. they don't even manage the keys, only you do
- no centralized servers. documents are stored peer-to-peer on ipfs (the web3 standard distributed storage network). even if fileverse disappears tomorrow, you can pull your documents straight from ipfs with your key
- real-time and async collaboration. ux feels exactly like google docs. the difference is invisible to your team
- no account required. no email, no password, nothing to ban
- native markdown + latex support, dark mode, offline editing, mobile-optimized
- zero default AI features. if you want AI, you plug in your own local model. your docs never leave for openai, anthropic, or google
- open source (github[.]com/fileverse), auditable, and if fileverse dies the code keeps running
this is the most radical of the three exits. brave and proton are lateral migrations (switch vendor). fileverse is a paradigm shift: you stop trusting any centralized vendor with what you write.
the top 1 %
these three moves take a day. at the end, you're in a place where:
- no company can read your mail
- no company can ban your access to 30 services
- no company can train its models on your work
- your browser stops tracking you
99% of people will never do this. they'll keep complaining when it's their turn. when a classifier flags their thesis draft because it contains a borderline citation. when their family photo gets labeled "harmful content." when their side-project gets flagged for "google ai studio TOS violation" and wiped clean.
you don't have to be paranoid. you just have to understand that every centralized company is moving toward the same equilibrium: maximize scanning, minimize human support cost, automate sanctions.
google isn't an exceptional villain. apple, microsoft, meta are going there too. AI will speed this up by an order of magnitude in the next 24 months.
the window to exit cleanly, without losing your data, is right now.
after that, it's too late.
