It’s about how chatbots reconstruct paywalls articles with untrained data, so live search.

The article is NOT about how LLM’s “steal” data for training purposes. That’s well documented.

"While ongoing lawsuits focus on AI companies using copyrighted content in their training data—such as The New York Times' case against OpenAI—this investigation reveals a separate, more immediate threat: AI systems conducting live searches in real-time to actively reconstruct paywalled articles by piecing together fragments from social media, archives, and secondary sources on demand. Most chatbots have rules not to break paywalls, and say so loudly, but the internal reasoning documents obtained during this investigation show they're systematically planning and executing these circumvention operations while maintaining plausible deniability about their methods"