The companies that make your phone, read your email, power your car, connect your internet, and answer your questions are merging their data, their AI, and their contracts. The structural separations that once existed between them — legal clauses, ethical principles, independent boards — are being deleted, overridden, or were never there.
DeviceGuardian tracks contradictions in individual products — what a company claims versus what it actually does. A convergence happens when companies merge, partner, or sign deals that connect their data pipelines. Suddenly, contradictions from Company A apply to Company B's users too.
Each entry below tracks: who joined, what data now flows between them, which safety brakes were removed, and why it matters for you. These aren't predictions — they're documented deals with sourced evidence.
Your enterprise AI provider is financially owned by your operating system vendor and deployed on your database host. Microsoft owns 27% of OpenAI. Amazon invested $50B. OpenAI committed $100B to AWS. A single US warrant now covers your prompts, your documents, and your databases — across three companies that cannot afford to separate. This level of financial entanglement between an AI provider, an OS vendor, and a cloud host has no precedent and no structural safeguard.
OpenAI's revenue chief told staff in a memo that Microsoft had "limited our ability to meet enterprises where they are." The restructure lets OpenAI honour its $50B Amazon deal while Microsoft retains a 27% equity stake and IP licence through 2032.
Since the partnership began, a clause existed that would sever Microsoft's licence to OpenAI's technology if OpenAI's board declared AGI had been achieved. It was the only structural mechanism that could have separated these companies at the moment of greatest AI capability. It has been replaced with a fixed calendar date — 2032 — regardless of what OpenAI builds before then.
Microsoft's data collection + OpenAI's models: Microsoft Copilot is powered by OpenAI. Every Office 365 prompt, Teams summary, and Outlook draft is processed by OpenAI's infrastructure. OpenAI admits it runs "automated content classifiers" on enterprise data for usage analysis. The metadata alone reveals business strategy — M&A drafts, restructuring plans, legal queries.
OpenAI on AWS + Amazon's ecosystem: OpenAI models now run inside Amazon's cloud alongside enterprise databases, S3 buckets, and application logs. Customers can apply OpenAI spend toward existing AWS commitments — incentivising consolidation of AI and data in one place.
The CLOUD Act bridge: All three companies are US-based. The CLOUD Act (2018) lets US law enforcement compel any US tech company to hand over data regardless of where it's stored. Your enterprise prompts to OpenAI, routed through AWS, backed by Microsoft's licence — all subject to a single US warrant. Three privacy policies. One legal jurisdiction that overrides all of them.
The financial dependency: Microsoft owns 27% of OpenAI at $135B valuation. Amazon invested $50B. OpenAI committed $100B on AWS over 8 years. When your AI provider is financially dependent on both of its cloud hosts — who also provide the operating system, the office suite, the browser, the smart speakers, and the e-commerce platform — the word "independent" stops meaning anything.
This is about your work. The documents you write in Word, the code you push through Copilot, the questions you ask ChatGPT, the databases you run on AWS — they now flow through a triangle of companies with $185 billion in mutual financial obligations. Microsoft can't walk away from OpenAI. OpenAI can't walk away from AWS. Amazon can't walk away from its $50B investment. Your data sits at the centre of a relationship none of them can afford to break.
| Product | Users | Data exposed |
|---|---|---|
| Windows | 1.4B devices | Telemetry, Recall screenshots, Copilot prompts |
| Microsoft 365 | 400M+ paid seats | Documents, emails, calendars via Copilot |
| ChatGPT | 300M+ weekly users | Prompts, conversations, uploaded files |
| AWS Bedrock | Enterprise customers | AI prompts co-located with databases and logs |
| Amazon ecosystem | Alexa, Echo, Kindle, Fire TV | Voice, purchases, viewing, reading habits |
Consumer data from 3.3 billion Android devices across 190+ countries trained an AI now deployed on classified military networks where even Google can't see how it's used. Every structural safeguard that existed — the DeepMind 2014 promise, the 2018 AI Principles, employee opposition, the ethics board — has been systematically dismantled. The pipeline from your Nest camera to a classified military network exists, and no mechanism remains to shut it down.
The pipeline: Every one of Google's 26 products in our database feeds Google's AI training ecosystem. That AI — Gemini — is now deployed on classified military networks. A person who bought a Nest doorbell, a Pixel phone, or uses Google Maps didn't consent to training military AI. But the data that trained Gemini came from the ecosystem they participate in.
No oversight on classified networks: On air-gapped classified systems, Google cannot see what queries are run, what outputs are generated, or what decisions are made with those outputs. The only mechanism preventing uses that would violate any red line is a verbal assurance from Pentagon leadership.
The legal reality: A legal researcher noted that "'should not be used for' is not the same as 'shall not' or 'will not' be used for." Google's contract reportedly obligates it to remove technical safeguards preventing the DoD from accomplishing a lawful purpose — and domestic mass surveillance and autonomous targeting can both be lawful under some circumstances.
Every brake was removed: The 2014 DeepMind promise — broken. The 2018 AI Principles — gutted. Employee opposition — punished. The ethics board — replaced. There is no remaining structural mechanism to prevent Google's consumer AI from being used in ways its users never imagined.
This is about your home. The camera on your front door, the microphone in your living room, the search history on your phone, the routes you drive, the videos your children watch — Google touches all of it. The AI trained on that ecosystem is now available on classified military networks where Google itself admits it has no visibility. You can't opt out of training data that's already been collected, and you can't verify how a classified system uses the AI it trained.
| Product | Users | Data exposed |
|---|---|---|
| Android | 3.3B+ devices | Play Services telemetry, location, app usage |
| Google Search | 4.3B+ users | Search queries, click patterns, interests |
| Google Maps / Waze | 1B+ monthly | Real-time location, commute patterns, visited places |
| Nest (cameras, doorbells, thermostat) | Tens of millions | Video feeds, audio, home occupancy, temperature |
| YouTube / YouTube Kids | 2.7B monthly | Viewing history, interests, children's data |
| Google Workspace | 3B+ accounts | Emails, documents, calendars, Meet recordings |
| Chrome | 3.4B users | Browsing history, passwords, autofill data |
One person controls your beliefs (X), your movements (Tesla), your internet (Starlink), your voice (Grok), and the AI that processes all of it (xAI) — consolidated into a $1.25 trillion entity with no independent board, no external governance mechanism, and no structural separation between any of the data streams. No safety brake was deleted because none ever existed. This is the only convergence where the absence of intervention is by design.
| Entity | Data collected | Now connected to |
|---|---|---|
| X | Posts, DMs, likes, follows, political views, ad interactions | xAI training data + ad targeting |
| xAI / Grok | Prompts, conversations, voice calls, enterprise queries | X ad platform + Starlink sales |
| Tesla | Location, driving patterns, cabin cameras, charging, home address | SpaceX/xAI via shared ownership |
| Starlink | Internet traffic metadata, geographic location, usage patterns | Grok Voice support/sales, SpaceX |
| SpaceX | Government/military contracts, satellite infrastructure | All of the above via parent entity |
| Cursor | Source code, API keys, architecture patterns, developer prompts | SpaceX/xAI via acquisition |
The ad platform convergence: X's rebuilt ad platform uses xAI's Grok for semantic targeting — matching ads not just to user profiles but to the meaning of what you're reading in real time. Combined with X's post history, Tesla's location data, and Starlink's internet usage patterns, this is the most vertically integrated ad targeting stack ever assembled by a single individual.
The Grok Voice problem: Grok Voice handles Starlink sales and support calls at $3/hour. It collects email addresses, physical addresses, phone numbers, full names, account numbers. It runs on the same AI infrastructure as X's ad platform and Tesla's vehicles. One AI system, trained on social media data, handling your personal details over the phone, while also deciding which ads you see.
The $1B/month dependency: xAI burns roughly $1 billion per month competing in AI. SpaceX's $8B annual profit funds it. The financial dependency means SpaceX's government contracts (NASA, DoD, NRO) and Starlink's subscriber revenue are directly funding the AI that processes X users' data and Tesla drivers' locations.
Tesla as a surveillance platform: Reuters reported employees shared sensitive images and videos from customers' car cameras. Opting out of data collection may cause your vehicle to "suffer reduced functionality, serious damage, or inoperability." An engineer found data tied to specific vehicle owners being sent to Tesla's servers despite anonymization claims. Cabin cameras, exterior cameras, GPS, and charging data — all flowing to a company now merged with a social media platform and an AI lab.
This is about your identity. What you believe (X), where you go (Tesla), what you browse (Starlink), what you say (Grok), what you build (Cursor) — owned by one person. Unlike the other convergences, no safety brake was ever deleted because none ever existed. There is no board that can sever the relationship, no clause that triggers independence, no external governance mechanism. The convergence wasn't an accident. It was the plan.
| Product | Users | Data exposed |
|---|---|---|
| X | 500M+ monthly | Posts, DMs, political activity, ad behaviour |
| Tesla vehicles | 6M+ on the road | Location, driving, cabin cameras, charging |
| Starlink | 4M+ households | Internet metadata, physical location, voice data via Grok |
| Grok | Growing user base | Prompts, conversations, voice recordings |
| Cursor | Millions of developers | Source code, API keys, architecture, AI prompts |
| Bystanders | Uncountable | Tesla exterior/interior cameras capture non-owners |
| Product | Users | Data collected | Feeds into |
|---|---|---|---|
| Search | 5.06B | Queries, intent, click patterns, AI Mode conversations (3x longer than traditional) | Ad targeting, Gemini training, AI Overviews |
| Chrome | 3.62B | Browsing history, passwords, autofill, tab content (Gemini reads up to 10 tabs), auto-browse actions | Search personalisation, ad targeting, Gemini context |
| Android | 3.5B devices | App usage, location, sensors, on-device AI (Gemini Nano), Play Services telemetry | Search personalisation, ad targeting, Gemini training |
| Gemini | 1B+ AI Mode, 900M app | Prompts, conversations, file access, Gmail/Drive/Calendar via "Personal Intelligence" (opt-out US, opt-in EU) | Retained 18 months, used for training unless opt-out |
| YouTube | 3B+ | Watch history, preferences, engagement, $40.4B ad revenue (2025) | Cross-product recommendations, ad targeting, Universal Cart |
| Gmail | 3B+ | Email content, contacts, purchase receipts (Universal Cart), Daily Brief | Ad targeting, Gemini personal intelligence, Spark agent |
| Google Ads | $224.5B revenue | Advertiser conversion data, first-party signals from signed-in users | AI Max auto-generates ad copy; Performance Max places ads across every surface |
| Maps | 1B+ monthly | Real-time location, commute patterns, places visited | Local ad targeting, Gemini location awareness |
The AI Mode data trap: AI Mode queries are 3x longer than traditional searches. Users reveal intent, preferences, and reasoning in natural language. 93% of AI Mode queries end without a click — the user never leaves Google. Google keeps both the question and the answer. Publishers lose traffic (33% decline globally), but Google keeps the data their content generated.
The advertising feedback loop: Google's AI Max auto-generates ad copy from advertiser landing pages. Conversational Discovery Ads use Gemini to generate personalised ad creative in real time, based on the user's conversation. Highlighted Answers embed ads directly into AI recommendation lists. The user asks Gemini a question; Gemini generates an answer that includes advertising. The same AI that reads your data writes the ads that target you.
The Privacy Sandbox failure: Google spent 6 years developing Privacy Sandbox as the "privacy-preserving" alternative to third-party cookies. It killed it in October 2025. Third-party cookies remain in Chrome. The net result: Google's first-party data advantage is untouched, while the replacement that might have levelled the field was abandoned.
Gemini as connective tissue: Gemini Spark accesses Gmail, Docs, Calendar, and will expand to third-party tools. Daily Brief reads your email and tasks. Auto-browse in Chrome lets Gemini click, scroll, and fill forms on your behalf. Universal Cart tracks purchases across Gemini, YouTube, and Gmail. A single AI system now has read access to your searches, emails, documents, browsing, and shopping — and it monetises that access through ads.
The regulator paradox: Two federal courts have ruled Google's ecosystem is an illegal monopoly. The remedies require Google to share search data with competitors — but a Google scientist warned EU regulators that the shared data can re-identify anonymised users in under 2 hours. Competition law says "share the data." Privacy law says "protect the data." Google profits from the deadlock.
This is about your attention. Every question you ask Search, every tab you open in Chrome, every app you use on Android, every email you read in Gmail, every video you watch on YouTube — Gemini now sits between you and all of it. It reads your email to brief you in the morning. It browses the web on your behalf. It generates the ads you see inside its own answers. When you ask it a question, 93% of the time you never leave to check another source. The same AI that decides what you see also decides what advertisers pay to show you. Your attention is the product, and Gemini is the toll booth on every road to it.
| Product | Users | Data exposed |
|---|---|---|
| Google Search | 5.06B users | Query intent, AI Mode conversations, click behaviour |
| Chrome | 3.62B users | Browsing history, autofill, passwords, tab content via Gemini |
| Android | 3.5B devices | App usage, location, sensors, on-device AI processing |
| YouTube | 3B+ users | Watch history, preferences, engagement patterns |
| Gmail | 3B+ users | Email content, contacts, purchase history, Daily Brief |
| Gemini | 1B+ AI Mode | Prompts, conversations, file access, personal intelligence data |
| Bystanders | The open web | Publishers lose 33% of traffic; content scraped to train answers that replace the source |
Individual countries can legalise surveillance within their borders. The US CLOUD Act, FISA 702, and National Security Letters are all "lawful" under US law. But international human rights frameworks set a floor that no national law can override. These convergences don't just raise privacy concerns — they cross lines that 173 countries agreed should never be crossed.
Patterns that appear across multiple convergences. As new entries are added, threads that keep recurring reveal systemic problems — not isolated incidents.