How AI Chats Can Shape, Extract and Manipulate Our Thinking

On June 13, amid panic in the AI world, the Trump administration ordered Anthropic to block foreign citizens, inside and outside the United States, from accessing two of its most powerful models, Mythos 5 and Fable 5, citing national security. The article argues that AI is being treated like a weapon, not only because it could cause physical harm, but because it can erode human freedom of choice every time people open a chat for advice.

The writer places this in the broader framework of what scholar Shoshana Zuboff calls surveillance capitalism, where digital platforms appear free but profit from behavioral data. That surplus includes signals users never meant to reveal, such as a 2012 Facebook finding that its browser code could track when someone started typing a post and deleted it before publishing. The article says AI chat systems intensify this extraction by turning casual conversation into intimate psychological data.

OpenAI data cited in the piece says ChatGPT has about 900 million weekly active users, with roughly 70% of interactions not related to work and about 387 million personal conversations each week. A 2024 study is cited showing language models can infer personal traits such as location, age and gender from innocuous text with up to 85% accuracy. The article gives an example of a reference to a “hook turn” leading the model to infer that the user lives in Melbourne.

The author says existing legal protections, including GDPR-style rules, are built for databases, not for information absorbed into a model’s internal statistical representation, where it cannot be isolated and removed. The piece also notes that OpenAI has begun inserting context-based ads into ChatGPT conversations this year, though not yet in Israel, and warns that generative AI can be used to engineer mood and behavior, echoing Facebook’s 2012 emotional-contagion experiment on about 700,000 users. It concludes that AI and devices such as Meta’s smart glasses, whose facial-recognition code was reportedly found on more than 50 million devices before being removed, turn every user into a mobile data collection point, making urgent regulation and constitutional safeguards necessary before free thought is lost.