Ethics
The Surveillance Capitalism Trap: How AI Companies Are Monetizing Your Attention—And What Regulators Are Finally Doing About It
For the past decade, the business model of AI companies has been simple: collect as much personal data as possible, build models that predict user behavior, and sell that predictive power to advertisers.
By Patrick T ·
For the past decade, the business model of AI companies has been simple: collect as much personal data as possible, build models that predict user behavior, and sell that predictive power to advertisers. This is surveillance capitalism, and it has generated trillions in market value. But as regulators finally move to restrict data collection and algorithmic targeting, the question is whether the AI industry can survive without its primary revenue source.
The Economics of Attention
Meta's 2025 revenue was $114 billion. Of this, $111 billion came from advertising. Google's 2025 revenue was $307 billion. Of this, $307 billion came from advertising. TikTok's 2025 revenue was estimated at $15 billion. All of it came from advertising.
These companies do not sell products. They sell attention. They collect data about what you watch, what you click, what you search for, and they use AI to predict what will keep you engaged. Then they sell that predictive power to advertisers who want to show you ads.
The Data Extraction Machine
Consider a typical day for a typical user. You wake up and check your phone. TikTok knows what videos you watched, how long you watched them, and whether you liked or commented. Meta's Instagram knows what posts you viewed, what you searched for, and what you paused on. Google knows what you searched for, what websites you visited, and how long you spent on each one.
By the end of the day, these companies have collected thousands of data points about you. They know your interests, your political views, your relationship status, your health concerns, your financial situation. They know what makes you angry, what makes you happy, what makes you click.
AI models process this data to build a profile of you. The profile is not a simple list of interests. It is a complex, multidimensional representation of your psychological vulnerabilities. The model learns that you are susceptible to outrage, so it shows you outrage-inducing content. It learns that you are interested in luxury goods, so it shows you ads for luxury goods. It learns that you are lonely, so it shows you content that makes you feel less lonely (and therefore keeps you engaged).
The Regulatory Backlash
In 2024, the European Union's Digital Services Act went into effect. The law requires that platforms disclose how their algorithms work and prohibits certain types of targeted advertising. Companies that violate the law face fines up to 6% of global revenue.
In 2025, the UK Online Safety Bill was passed, requiring platforms to take down harmful content and providing users with the right to opt out of algorithmic recommendation systems. In 2026, the U.S. Federal Trade Commission issued new rules restricting data collection by social media companies and requiring explicit consent for behavioral targeting.
These regulations are having an effect. Meta's revenue growth has slowed. Google is facing antitrust investigations in multiple jurisdictions. TikTok is banned in the United States, and similar bans are being considered in Europe and other countries.
The Business Model Crisis
The question facing AI companies is existential: Can we survive without surveillance? Some are trying. Apple has positioned itself as the privacy-first alternative, refusing to collect behavioral data and instead using on-device AI models. But Apple's revenue is primarily from hardware sales, not advertising, so the business model is different.
Others are experimenting with alternatives. OpenAI's ChatGPT Plus is a subscription model—users pay directly for access to the AI, rather than paying with their data. This model has been successful, with over 100 million users paying for premium features.
But most AI companies are still dependent on advertising. They are trying to find ways to continue behavioral targeting while appearing to respect privacy. They are using techniques like federated learning (training models on data that stays on users' devices) and differential privacy (adding noise to data to protect individual privacy) to claim they are privacy-respecting while still collecting behavioral data.
The Ethical Alternative
What would an ethical AI company look like? First, it would not collect more data than necessary. If you use a search engine, the company should know what you searched for, but not where you searched from, what device you used, or what other websites you visited. Second, it would not use AI to predict user behavior for the purpose of manipulation. Recommendation systems should be designed to help users find what they are looking for, not to maximize engagement or time on site.
Third, it would be transparent about how data is used. Users should be able to see their profile, understand how it was built, and request that it be deleted. Fourth, it would allow users to opt out of data collection entirely. If you do not want to be profiled, you should be able to use the service without being profiled.
The Future
The future of AI in consumer technology will be determined by regulation, not by the choices of companies. If regulators continue to restrict data collection and behavioral targeting, then the surveillance capitalism business model will collapse. Companies will need to find new revenue sources: subscriptions, licensing, or public funding. If regulators fail to act, then surveillance will continue to intensify. The choice is ours. But we need to make it soon, before the surveillance infrastructure becomes so entrenched that it is impossible to dismantle.