Artificial Intelligence (AI) > AI surveillance and data collection
AI surveillance and AI data collection, explained
AI surveillance and data collection aren’t limited to the tools you intentionally use. They occur through ads, social media, the tools you use at work, and even in public spaces. Here's a closer look into AI surveillance and data collection — and what it means for your privacy.
Definitions of AI surveillance and AI data collection
AI surveillance refers to the use of AI to watch people — whether that's recognizing faces in a crowd, detecting unusual behavior in public spaces, or monitoring what employees do at work. The goal is typically some form of observation, identification, or control.
AI data collection is the foundation of what makes surveillance not just possible, but extremely effective. Information gathered from apps, devices, services, and platforms are all used to build profiles of people, including their habits, preferences, and behaviors. Data collection doesn't always involve just watching you in real time; it can happen quietly in the background too. And while this data can power surveillance, it also serves other purposes, including training AI models and targeting ads.
| Examples of AI surveillance | Examples of AI data collection |
|---|---|
| Facial recognition cameras in malls and airports | Conversations logged in chatbots |
| Workplace monitoring of keystrokes and screen activity | Apps track your location in the background |
| Smart devices that can detect movement or voices | Websites record clicks and browsing habits |
Surveillance and data collection — then and now
Before AI, large-scale surveillance still required people to manually watch screens, sift through records, or cross-reference datasets. AI changed that by making it faster and cheaper to scan images, transcribe conversations, flag patterns, and combine disparate data sources into a single profile.
The result isn't just more data collection; it's more inference and loss of privacy. Even if you never explicitly share your interests, habits, or beliefs, AI can often deduce them from location trails, browsing patterns, and social behavior. That's what makes AI surveillance and data collection more powerful — and more dangerous — than ordinary tracking alone.
AI surveillance for control
When used for control, AI turns public spaces, roads, workplaces, and social platforms into environments where mass surveillance becomes not just easy but routine.
Public-space monitoring
In cities worldwide, AI surveillance takes the form of facial recognition, smart CCTV, crowd analytics, and automated license plate readers. These systems can identify individuals, track movements, and detect anomalies in real time.
The practical consequence is the erosion of anonymity in public life. Technologies that match faces, vehicles, or devices across locations can record and reconstruct ordinary movement — even when a person has done nothing wrong.
In law enforcement and civic monitoring
Police and civic agencies increasingly combine camera feeds, license plate records, public databases, and social media activity into unified investigative systems. AI can then scan vast volumes of footage and posts to surface patterns, flag individuals, or identify connections between people.
What makes this especially concerning is how this footage and data can be used to infringe on privacy rights. For example, protest attendance, social networks, or everyday movements can become data points, which can cause people to self-censor or avoid public gatherings out of fear of being watched.
In the workplace
Modern workplace monitoring has evolved from simple time-tracking to deep behavioral analysis. Systems now log keystrokes, track mouse movements, monitor screen activity, and use AI to analyze communication sentiment. One example involved Meta’s plan to capture employee activity such as clicks and keystrokes for AI training purposes, while Walmart and Starbucks were reported to be using AI tools to monitor employee communications(new window).
While employers often frame these tools as productivity measures, continuous behavioral monitoring raises serious privacy concerns — and in some jurisdictions, it crosses a legal line. The use of certain productivity-tracking systems violates EU privacy laws and is illegal in countries like Italy.
On social platforms
Social media is one of the most data-rich surveillance environments, where identity, relationships, communication, and behavior all converge. AI analyzes posts, photos, likes, and network structures to infer everything from interests to social connections. Because these algorithms operate behind the scenes of recommendation engines and moderation tools, users often remain unaware of the depth of inference occurring from their ordinary activity.
Surveillance capitalism and AI profiling
Surveillance capitalism is the business model of collecting everything you do online — and turning that data into profit. AI strengthens this by making it easier to combine data from apps, websites, platforms, and data brokers into highly detailed behavioral profiles used to shape recommendations and serve targeted ads.
Unlike state surveillance (which is usually tied to control, enforcement, or security), surveillance capitalism is driven by profit. The goal is to understand user behavior well enough to influence what they see, do, and buy.
On phones and apps
Mobile devices are uniquely revealing due to the sheer amount of personal data they contain. Apps frequently collect advertising IDs, precise location data, IP addresses, and usage telemetry — often without you knowing. Over time, that data can expose where you live, work, shop, seek medical care, and worship.
Across websites and adtech networks
Every time you visit a website, pixels and cookies silently record your activity, including how long you stayed and where you went next. That information gets passed along a network of adtech companies, each adding more detail to a profile of your behavior. By the time an ad appears on your screen, your data may have already traveled through dozens of companies you've never heard of.
But beyond just selling your products, this same infrastructure can also support state-level surveillance. US government agencies have also purchased location data from internet advertising brokers, illustrating how commercial tracking systems can also be repurposed for enforcement and investigation.
Through data brokers
Data brokers create detailed profiles about individuals through many sources, including social media, location, and online activity, then sell that data to third parties, often without the consent of said individuals. These corporations operate largely out of sight, but recent enforcement actions reveal how widespread the problem is. In January 2025, the FTC banned Mobilewalla(new window) from selling sensitive location data after finding it had illegally collected and sold data on millions of people without their knowledge.
In AI tools and assistants
AI platforms themselves are part of this economy. Depending on the tool you use, prompts, uploaded files, conversation history, and technical metadata may be stored, reviewed, or used to train models. As people increasingly use AI for work, planning, and sensitive tasks, all your conversations can be collected, monitored, and monetized.
How much data does AI collect?
The scope of AI data collection is substantial and growing. OpenAI's GPT-3 model, for example, was trained on roughly 300 billion tokens of text, while newer models are estimated to use trillions of tokens. AI data collection can encompass everything from biometrics like your face, voice, and fingerprints, to browsing habits, beliefs, and behavioral patterns.
If you use an AI platform, the amount and type of data collected depends on your plan and usage. Here are the types of data collected by popular AI tools, according to their privacy policies.
| Platform | Types of data collected |
|---|---|
| ChatGPT | Prompts, uploaded content, account information, usage data, and technical data |
| Gemini | Text, voice, files, photos, feedback, device information, and connected-service context |
| Claude | Conversations, files, feedback, account details, and technical data |
| Grok | Prompts, X-integrated data (including X posts and interactions), location, and technical data |
| Meta AI | Prompts, conversations, account and profile information, and cross-app behavioral from Facebook, Instagram, Messenger, WhatsApp |
| Copilot | Prompts, conversations, interactions, files, personalization settings, voice input, and usage data |
How to reduce AI surveillance and data collection
While it's not feasible to eliminate every form of monitoring, you can take steps to reduce how much data is being collected about you:
- Avoid entering sensitive personal, financial, legal, medical, or work information into AI platforms or making them publicly available.
- If you use an AI tool or software, carefully review its privacy policy and default settings.
- Limit app and browser permissions, especially for location, microphone, camera, photos, and contacts.
- Be cautious with connected accounts, since email, calendars, cloud files, and messaging integrations can expand what an AI tool can access.
- Choose AI tools that do not monitor, log, or use your chats for training by default.
Switch to a private AI assistant
Lumo is designed for people who want an AI assistant they can trust. Lumo never trains on your chats nor retains any logs, ensuring your data is completely private. Not even Proton can access it.
Frequently asked questions about LLMs and AI security
- Is AI data collection always used to train models?
- Can I stop companies from using my data for AI training?
- Is there an AI tool that doesn’t collect data?
