Anthropic

For the philosophical and cosmological concept, see Anthropic principle.

Anthropic PBC
ऐन्थ्रोपिक्

Industry	Artificial intelligence
Founded	January 26, 2021 (2021-01-26)
Founder(s)s	Dario Amodei Daniela Amodei Jared Kaplan Chris Olah Jack Clark Ben Mann Sam McCandlish Tom Brown
Headquarters	500 Howard Street, San Francisco, California, U.S.
Key people	Template:Ubil
Products	Claude Claude Code Claude Cowork Claude Design Bun
Number of employees	2,500 (2026)
Website	{{URL|example.com|optional display text}}

Anthropic PBC is an American artificial intelligence company headquartered in San Francisco, California. It develops a family of large language models (LLMs) known as Claude, with a strong emphasis on AI safety and responsible deployment. Anthropic was founded in 2021 by former OpenAI employees, notably the siblings Daniela Amodei (president) and Dario Amodei (CEO). The company remains privately held; as of May 2026, its estimated valuation stood at 965 arab US$, positioning it as the world's most valuable pure-play AI enterprise.

Projects

U.S. Military and Intelligence Applications

In November 2024, Anthropic entered into a partnership with Palantir Technologies and Amazon Web Services to deliver the Claude model to U.S. intelligence and defence agencies. In June 2025, the company introduced a specialised "Claude Gov" variant tailored for national security needs. According to reports, this model was already in use across several U.S. national security agencies by mid-2025.

As of February 2026, Anthropic's collaboration with Palantir enabled Claude to become the sole AI model deployed in certain classified U.S. missions. In July 2025, the United States Department of Defense (DoD) awarded Anthropic a contract worth up to $200 million for AI capabilities in military applications, alongside similar arrangements with Google, OpenAI, and xAI.

According to The Wall Street Journal, the U.S. military employed Claude during its 2026 operation in Venezuela that resulted in the capture of President Nicolás Maduro. The precise manner of its application remains undisclosed. The intervention led to the deaths of 83 individuals. Anthropic's acceptable use policy explicitly bars the direct employment of Claude for domestic surveillance or in fully autonomous weapons systems. White House officials noted that these restrictions had prevented certain federal contractors, including those working with the FBI and Secret Service, from utilising the model. This stance has generated tensions with the Pentagon and the Trump administration.

On 24 February 2026, Defence Secretary Pete Hegseth issued an ultimatum, threatening to exclude Anthropic from the DoD supply chain by 27 February unless the company agreed to unrestricted use of Claude. He indicated the possibility of invoking the Defence Production Act to enforce compliance. On 26 February, Anthropic declined to remove its AI safeguards. The following day, President Donald Trump directed U.S. government agencies to cease using Anthropic-developed models, characterising the move as retaliatory.

Two days later, reports emerged that the U.S. military had continued to access Claude through Palantir's Maven Smart System during initial strikes in the 2026 Iran conflict. The Pentagon subsequently launched an investigation into whether Claude contributed to a U.S. strike on an Iranian girls' school that resulted in over 170 fatalities, predominantly children. On 5 March, the Pentagon formally designated Anthropic a supply chain risk. In response, the company stated it had "no choice but to challenge" the decision in court.

In March 2026, executives from Google, Amazon, Apple, and Microsoft publicly backed Anthropic's legal challenge against the DoD. On 11 March, a Pentagon memorandum indicated that Anthropic's AI tools could be employed in exceptional national security circumstances, extending beyond an initial six-month phase-out period. On 26 March, the presiding federal judge granted a temporary injunction against the Pentagon's measures, observing that the actions appeared to constitute "classic First Amendment retaliation."

Project Panama

In January 2026, court filings from a 2024 class-action copyright lawsuit against Anthropic were unsealed, exposing details of the company's secret "Project Panama" initiative. Internal planning documents characteristically describe Project Panama as Anthropic's ambitious "effort to destructively scan all the books in the world".

To execute this project, Anthropic procured millions of used books from online retailers including Better World Books. The operational procedure involved removing book spines and scanning their pages to train their Claude AI model. Following the scanning process, the paper materials were recycled. The company engaged Tom Turvey, a key figure in the creation of Google Books, to oversee this operation. Internal documentation indicated that Anthropic deliberately maintained secrecy about Project Panama, with planning documents explicitly stating the company's desire that "it to be known that [it was] working on this".

In a significant legal development, Judge William Alsup ruled that the destruction of legally purchased books constituted fair use. This ruling contrasted sharply with Anthropic's earlier practice of using pirated copies for similar purposes.

Legal Issues

On 18 October 2023, Anthropic faced legal action from Concord, Universal, ABKCO Music, and other music publishers, who filed a lawsuit alleging "systematic and widespread infringement of their copyrighted song lyrics". The plaintiffs contended that Anthropic had utilised copyrighted material without permission in the form of song lyrics. The legal action sought damages of up to 1,50,000 US$ for each infringed work, citing violations of copyright laws.

The plaintiffs substantiated their allegations with specific examples, demonstrating how Anthropic's Claude model had reproduced copied lyrics from popular songs such as Katy Perry's "Roar" and Gloria Gaynor's "I Will Survive". Notably, the lawsuit alleged that even when provided with prompts that did not explicitly mention song titles, the model would still respond with modified lyrics based on original copyrighted works. In response, Anthropic, on January 16, 2024, argued that the music publishers had not suffered unreasonable harm and characterised the examples cited by the plaintiffs as mere software bugs.

In August 2024, a separate class-action lawsuit was initiated against Anthropic in California, this time alleging copyright infringement through the feeding of pirated copies of authors' works into its large language models. The suit specifically named participants Kirk Wallace Johnson, Andrea Bartz, and Charles Graeber as affected authors. On On 23 June 2025, the United States District Court for the Northern District of California delivered a partial summary judgment in favour of Anthropic, ruling that the use of digital copies of the plaintiffs' works for training LLMs constituted fair use. However, the court simultaneously found that Anthropic had utilised lakhs of pirated library copies, and that this specific use could not be classified as fair use. Consequently, the case was ordered to proceed to trial concerning the pirated copies used to create Anthropic's central library and the resultant damages. In September 2025, Anthropic reached a settlement with the authors, agreeing to pay 1.5 arab US$, which amounted to 3,000 {{USD} per book plus interest. This proposed settlement, pending judicial approval, represents the largest copyright resolution in United States history.

In June 2025, Reddit initiated legal proceedings against Anthropic, accusing the company of "unlawful and unfair business acts". The complaint alleged that Anthropic had violated Reddit's user agreement by training its AI models on users' personal data without obtaining their consent.