Sen. Chris Murphy (D-Conn.) accused artificial intelligence companies of undermining human creativity by acquiring rare books, digitizing their contents, and disposing of the physical originals.

Murphy Warns of AI Book Training

On Tuesday, Murphy posted on X a clip discussing what he described as a massive campaign by AI companies to acquire and digitize scarce books. He argued that while these companies claim their technology enhances human creativity, they are quietly destroying it.

In the video, Murphy alleged that AI companies are "literally buying up millions of books," including rare works with only one or a few remaining copies. He stated that after scanning these works, the companies dispose of them.

Anthropic’s Alleged Rare Book Project

Murphy specifically pointed to Anthropic and its "Project Panama," an effort documented in court filings. He described the project as involving the purchase of millions of physical books, the removal of their bindings, scanning of pages, and disposal of the originals.

Murphy raised concerns that such practices put scarce and potentially rare works at risk. He also claimed that AI companies are particularly interested in books created before 2022 because they represent work produced without AI assistance.

"These were wholly human-created works," he said, calling the alleged practice "especially perverse."

Murphy argued that destroying rare physical works could make it harder for future generations to access examples of original human creativity and distinguish them from AI-generated content.

He called for congressional action, suggesting lawmakers establish "common sense guard rails" around AI development. "The pace of this development, the pace of this innovation is frightening," he said, arguing that AI companies should operate "a little bit slower and a little bit more responsibly."

Industry Perspectives on AI and Creativity

Other industry leaders offered contrasting views on the impact of AI on work and creativity.

Kevin O’Leary stated that companies failing to adopt AI risk falling behind competitors. He noted that businesses increasingly use the technology for content creation, advertising, customer acquisition, and management. O’Leary said AI could significantly reduce production and creative costs while improving output, but emphasized the importance of human editing.

Nvidia Corp CEO Jensen Huang told parents not to focus on finding "AI-proof" fields for their children. He argued that skills such as storytelling, creativity, communication, and judgment would remain valuable. Huang encouraged students to learn how to use AI to enhance their work, stating that automation could shift workers toward higher-level tasks requiring human judgment.

OpenAI CEO Sam Altman predicted that AI could surpass human intelligence by 2030, potentially making scientific discoveries humans could not make independently. He estimated that 30% to 40% of economic tasks could eventually be performed by AI. Altman urged people to develop adaptability, resilience, and the ability to keep learning as technology reshaped jobs.

Anthropic has hired former Google TPU leader Amir Salek to lead its custom silicon efforts. The appointment marks a significant step in the AI developer's strategy to build proprietary hardware infrastructure and reduce dependence on scarce Nvidia GPUs.

Strategic Shift

The hiring signals Anthropic's move beyond software-only operations. By bringing in expertise from a major chip designer, the company aims to optimize its computing stack for large language models. Salek will report to Head of Compute James Bradbury and support the development of compute infrastructure.

Leadership Background

Prior to joining Anthropic, Salek was a senior managing director at Cerberus Capital Management. At Google, he served as senior director of engineering, delivering the first seven generations of TPU solutions to Google Data Centers. He also founded and scaled Nvidia's System-on-a-chip (SoC) organization and previously worked as a chip lead at PMC-Sierra.

Hardware Strategy

Anthropic is building out its internal chip-design team to develop custom processors for its Claude artificial intelligence models. The effort expands the company's multi-chip strategy, which includes hardware and cloud partnerships with Amazon Web Services, Google, Nvidia, and AMD.

Developing a cutting-edge AI chip can cost roughly $500 million. This expense covers recruiting specialized engineers, designing advanced architectures, and ensuring chips can be manufactured at scale without costly production failures.

Existing Partnerships

Anthropic maintains deep ties with Amazon, which has invested billions in the startup. In April, Amazon announced that Anthropic would spend more than $100 billion over the next 10 years on AWS Technologies. Additionally, Anthropic has bought $250 million worth of chips from U.K.-based Fractile and plans to expand that contract. The company has also signed capacity deals with Riot Platforms and Volta Infra Holdings Ltd.

What the Numbers Show

The financial scale of Anthropic's hardware ambitions is evident in its capital commitments. With a $100 billion commitment to AWS over a decade and a $250 million purchase from Fractile, the company is diversifying its compute sources. The $500 million cost estimate for developing a single cutting-edge AI chip highlights the significant capital allocation required for proprietary hardware, suggesting that internal chip design is a long-term infrastructure play rather than an immediate operational fix.