 | |
| Industry | Artificial intelligence |
|---|---|
| Founded | January 26, 2021[1] |
| Founder(s)s |
|
| Headquarters | 500 Howard Street,[6] San Francisco, California, U.S. |
Key people | Template:Ubil |
| Products | |
Number of employees | 2,500 (2026)[7] |
| 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".[8]
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.[8][9] 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.
References
- ↑ Cite error: Invalid
<ref>tag; no text was provided for refs namedincorporation - ↑ Lin, Belle. "Google and Anthropic Are Selling Generative AI to Businesses, Even as They Address Its Shortcomings". WSJ. Archived from the original on 10 April 2024. Retrieved 11 April 2024.
- ↑ 3.0 3.1 3.2 Ian Thomas (10 June 2025). "CNBC Disruptor 50". CNBC. Retrieved 16 February 2026.
- ↑ Nellis, Stephen (9 May 2023). "Alphabet-backed Anthropic outlines the moral values behind its AI bot". Reuters. Archived from the original on 5 June 2023. Retrieved 4 June 2023.
- ↑ Cite error: Invalid
<ref>tag; no text was provided for refs namedRoose - ↑ Waxmann, Laura (22 September 2025). "Growing tech firm Anthropic expands S.F. presence with large office lease in SoMa". San Francisco Chronicle. Archived from the original on 2 February 2026.
- ↑ Waxmann, Laura (30 January 2026). "Emerging tech titan Anthropic to take over an entire office building in downtown S.F." San Francisco Chronicle. Archived from the original on 10 March 2026. Retrieved 31 January 2026.
- ↑ 8.0 8.1 Schaffer, Aaron; Oremus, Will; Tiku, Nitasha (27 January 2026). "Inside an AI start-up's plan to scan and dispose of millions of books". The Washington Post. Archived from the original on 25 February 2026. Retrieved 2 February 2026.
- ↑ Edwards, Benj (25 June 2025). "Anthropic destroyed millions of print books to build its AI models". Ars Technica. Archived from the original on 26 June 2025. Retrieved 26 June 2025.