The BBC is threatening legal action against AI search engine Perplexity, accusing the US start-up of copyright infringement in a formal letter demanding it cease using its content. This move marks a major escalation in the global battle between media organizations and AI firms over data scraping.
The broadcaster alleges Perplexity reproduces its journalism verbatim and causes reputational harm through inaccurate summaries. In a sharp rebuttal reported by the Financial Times, Perplexity called the claims “manipulative” and accused the BBC of misunderstanding technology and law.
The AI firm further contended the action was an attempt to protect Google’s market dominance, stating “[The claims] also show how far the BBC is willing to go to preserve Google‘s illegal monopoly for its own self-interest.” This defiant stance sets the stage for a high-stakes legal fight over the future of AI and online content.
A Widening Legal War on AI
Perplexity is increasingly finding itself in the legal crosshairs of major media outlets. The BBC’s threat joins a growing list of legal challenges, including an October 2024 cease and desist letter from The New York Times and a subsequent copyright lawsuit from News Corp. This pattern of litigation is not unique to Perplexity. In April 2025, Ziff Davis, a digital media giant, filed a lawsuit against OpenAI, making similar claims of infringement and reputational damage.
The conflict is now a sprawling, global affair. A multi-district litigation in the US combines twelve separate cases from news organizations and authors against OpenAI and Microsoft. The issue also extends to book publishers, with the Federation of Indian Publishers suing OpenAI in January 2025 over its use of copyrighted literary works. These lawsuits collectively challenge the foundational practices of an industry that has, until now, largely operated under the assumption that public web data is free to use for training.
Fair Use on Trial
The core legal defense of the AI industry rests on the principle of “fair use,” arguing that training models on vast datasets is a transformative act. However, this defense is facing intense scrutiny. The argument is being directly tested in court for the first time in the context of generative AI in the case of Thomson Reuters v. ROSS Intelligence. Meanwhile, new evidence is undermining the claim that AI models merely learn concepts. A recent academic study revealed that Meta’s Llama 3.1 model had effectively memorized 42% of Harry Potter and the Sorcerer’s Stone, allowing it to reproduce the text verbatim.
This research lends weight to the argument that AI models contain infringing copies of protected works. Stanford law professor Mark Lemley, a co-author of the study, stated the findings show what “the law would call a copy of part of the book in the model itself.” This perspective was bolstered by the U.S. Copyright Office, the U.S. Copyright Office, which concluded in May that AI training is unlikely to be considered fair use if the model’s output harms the market for the original work. The BBC’s letter echoes this, claiming Perplexity’s tool directly competes with its own services.
To Sue or To Sign: The Publisher’s Dilemma
As the legal battles intensify, a dual-track strategy has emerged within the media industry: sue some, sign with others. While The New York Times is in a high-profile lawsuit with OpenAI, it simultaneously struck a major content licensing deal with Amazon in May 2025. This pragmatic approach is becoming more common as publishers seek compensation. OpenAI already secured licensing agreements with many publishers, including The Associated Press and Axel Springer, while Microsoft has deals with Reuters and the Financial Times. John Ridding, CEO of the FT Group, explained the rationale, stating it is “right, of course, that AI platforms pay publishers for the use of their material.”
This contrasts with the approach of other content owners who are proactively blocking AI firms. In October 2024, publisher Penguin Random House updated its copyright policy globally to explicitly forbid the use of its books for AI training. Perplexity, for its part, has attempted to find a middle ground by offering a revenue-sharing program, though it has struggled to gain traction with the largest media conglomerates, who increasingly favor direct licensing deals.
Beyond Copyright: A Question of Trust and Security
Scrutiny of Perplexity now extends beyond copyright into the technical integrity and security of its products. A security audit of Perplexity’s Android app by the firm Appknox flagged ten major vulnerabilities. The report found “hardcoded secrets” like API keys embedded directly in the app’s code, which could allow attackers to clone the application for phishing attacks.
The audit also found the app was vulnerable to “task hijacking” and lacked SSL pinning, a crucial security measure that prevents interception of communications. These findings raise serious questions about the company’s development priorities. Raghunandan J, head of R&D at Appknox, noted that in the race to innovate, “organizations today are wanting to come out with their applications and their updates faster, thereby ignoring their security.”
For the BBC, which is funded by the public and operates under strict editorial guidelines, the risk of its content being misrepresented by an insecure or inaccurate AI platform represents a fundamental threat to the trust it has built with its audience.
This multi-front battle—spanning legal, ethical, and technical domains—places Perplexity at the center of the AI industry’s broader reckoning. The BBC’s entry into the fray adds the weight of a major public service broadcaster to the chorus of creators demanding accountability. The outcome of these conflicts will not only determine the future of Perplexity but will also establish the rules of engagement between AI and the world of content for years to come.
