Bright Data says publishers must use logins to block AI agents

Bright Data’s stance on AI and public data

Bright Data, a provider of web data infrastructure, has advised publishers to put content behind a login if they want to prevent AI agents from accessing it. Jennifer Burns, the company’s Senior Director of PR and Communications, stated that any content accessible without signing in is considered public and therefore fair for AI use. She added that a paywall qualifies as a login, and Bright Data would not scrape content behind one.

Burns framed the distinction between public and non-public data as a matter of transparency. "If you make it accessible to any person, it should also be accessible to AI. There, you really can’t pick and choose who is going to be able to view that data and at what scale," she said. Anything short of this distinction, she argued, amounts to "gatekeeping."

Legal disputes and market growth

Bright Data’s data collection practices have led to legal conflicts with Meta and X. According to Burns, the company fought these cases in court, and rulings have so far upheld the legality of its operations. The outcomes align with Bright Data’s position that public data is fair for AI training and other uses.

The company’s customer base has grown from 20,000 to 50,000 over the past three and a half years, with staff numbers rising from 250 to around 600. Bright Data serves 14 of the top 20 large language models (LLMs) and handles 100 million agent interactions daily. Burns also mentioned that robotics companies began approaching Bright Data for public data to train on in late 2025.

AI agents in journalism and enterprise

Bright Data provides its tools to journalists for free through a dedicated programme, reflecting its commitment to supporting investigative work. Burns herself uses five AI agents in her role, which she said enhances rather than replaces her work. "It doesn’t mean that I am less relevant. It makes me more relevant because I have to manage those agents. I can’t let them run without a human in the loop," she noted.

She also cautioned against overusing the term "AI," saying, "Don’t throw around the word AI for the sake of AI. Everything is AI." The company has expanded into the "context layer," which involves providing live data—such as real-time prices—to AI agents for dynamic decision-making.

Legal and market context for data scraping

Bright Data’s position mirrors ongoing debates about fair use and AI training data. Earlier reporting from Lazyfounder highlighted the Wikimedia Foundation’s amicus brief in The New York Times’ lawsuit against OpenAI, which argued for a balanced approach to fair use. The foundation warned that extreme interpretations—either overly restrictive or permissive—could harm innovation or create a closed web.

Reddit’s recent shift to require logins for access, driven partly by declining search traffic, also reflects the tension between openness and control over data. While Reddit’s move was framed as a revenue strategy, it underscores how platforms are rethinking access to public content.

What this means

Lazyfounder analysis — our interpretation, not reported fact.

Bright Data’s stance sharpens the divide between public and non-public data in the AI economy. For publishers, the message is clear: if they want to block AI agents, they must gate their content behind logins or paywalls. This isn’t just a technical hurdle—it’s a strategic choice with legal and commercial implications.

The company’s legal wins against Meta and X suggest that, for now, courts are siding with broad access to public data. That could embolden other AI data providers, even as publishers like The New York Times push back through litigation. Founders building AI tools—or relying on scraped data—should note that the bar for what counts as "public" may be higher than they assume.

For operators outside the publishing world, the story is a reminder that data access isn’t just about technology; it’s about compliance and risk. If your business model depends on web scraping, Bright Data’s growth and legal record offer some reassurance. But the debate over what’s truly public—and who gets to decide—is far from over.

Related on Lazyfounder

• Wikimedia Foundation urges balanced fair use in NYT’s copyright lawsuit against OpenAI
• Wikimedia Foundation argues AI training is fair use but RAG responses are not in OpenAI lawsuit
• Reddit to require logins for access after search traffic declines
• Meta
• Reddit
• The New York Times