Anthropic

As of May 2026, Anthropic was valued at $965 billion. ⁴

Background

Anthropic was founded as a public benefit corporation in January 2021 by siblings Dario Amodei and Daniela Amodei, both of whom worked as executives at OpenAI, but left with five colleagues over concerns about AI safety. ¹ The name “Anthropic” is meant to evoke “anthropocentric,” or the idea that the company is concerned with humanity. ⁵

In May 2021, Anthropic raised $124 million in funding from Skype co-founder Jaan Tallinn, Infotech founder James McClave, Facebook co-founder Dustin Moskovitz, ex-Google CEO Eric Schmidt, and the Center for Emerging Risk Research, among others. ⁵

In February 2023, Anthropic raised $300 million in funding from Google. ⁶

In March 2023, Anthropic publicly released Claude, its large language model (LLM). Anthropic claimed that Claude operated similarly to OpenAI’s ChatGPT but was easier to converse with and “less likely to produce harmful outputs.” Initially, Claude had no ability to access the internet and was “contained,” though users were later given the option to give Claude access to specific websites. ⁶

In November 2025, Microsoft and Nvidia announced a commitment to invest $15 billion in Anthropic, which the AI company would then use as part of a $30 billion deal to buy computing power from Microsoft Azure running Nvidia systems. The plan was seen as an “AI alliance” against OpenAI, the leading AI company. ⁷

AI Safety

Anthropic co-founders Dario Amodei and Daniela Amodei left OpenAI to launch Anthropic over concerns about OpenAI’s safety protocols. They have emphasized that they do not oppose the commercialization of AI, but they believe that OpenAI puts too little emphasis on mitigating the dangers of AI, particularly extreme risk scenarios that may come from artificial general intelligence (AGI), a theoretical point at which artificial intelligence technologies surpass human intelligence in nearly all capacities and can further develop their own intelligence capabilities, thereby possible expanding the power of AI beyond human control. The Amodeis have claimed that OpenAI has scaled its AI technologies successfully but recklessly, without accounting for proper controls on AI behaviors or doing sufficient investigations into the inner workings of its AIs. ^{2 3}

Anthropic’s AI development is based on a concept called “Race to the Top,” in which Anthropic attempts to inspire more ethical AI development in other AI companies through its own conduct. Anthropic contrasts this model with Open AI’s “race to the bottom,” wherein AI companies compete against each other by improving their AIs as quickly as possible for commercial gain while increasing global AI risk. ^{2 3}

Anthropic’s principles limit Claude’s behavior through “Claude’s Constitution,” which places more restrictions on how Claude “thinks” and on how users can interact with Claude than with other LLMs. ⁸

In September 2025, Anthropic declared that it would stop selling its products to organizations that are majority-owned by Chinese, Russian, Iranian, or North Korean citizens due to concerns over national security. ⁹

In November 2025, Anthropic acknowledged that hackers employed by the Chinese government had bypassed Claude’s safeguards and used the AI to perform cyberattacks against 30 organizations. ¹⁰

In February 2026, the U.S. Department of Defense (DoD) demanded that Anthropic remove its internal restrictions on the use of its AI for domestic surveillance and weapons systems. After Anthropic refused, the second Trump administration banned all federal agencies from contracting with Claude. Later in the same day, OpenAI announced that ChatGPT would be used by the DoD. ¹¹ Anthropic subsequently sued the federal government; as of June 2026, the lawsuit is ongoing. ¹²

In June 2026, the U.S. government ordered Anthropic to block all access for foreigners to its Claude Fable 5 AI due to national security concerns. Soon after, Anthropic took Claude Fable AI offline to improve its security measures. ¹²

Criticism

Anthropic’s safety-oriented approach to artificial intelligence has been criticized for being overly restrictive. In October 2025, second Trump administration AI czar David Sacks and White House AI policy advisor Sriram Krishnan accused Anthropic of engaging in fear-mongering and limiting the competitiveness of American AI companies. Sacks also accused Anthropic of attempting “sophisticated regulatory capture” by lobbying for more government regulations on AI development that would hinder the growth of rival AI companies. ^{13 14}

Controversies

In September 2025, Anthropic agreed to pay $1.5 billion to settle a class-action lawsuit from a group of authors who alleged that the company illegally used copyrighted material to train Claude. The settlement paid $3,000 for each of the 500,000 books that the company allegedly pirated. Similar copyright lawsuits were concurrently filed against other artificial-intelligence companies, but Anthropic was the first to settle. ¹⁵ As of March 2026, Anthropic was continuing to fight similar copyright lawsuits filed by record labels regarding Claude’s alleged use of copyrighted music. ¹⁶

In June 2025, Reddit announced that it was suing Anthropic after revelations that the AI company used bots to scrape Reddit’s forums for Claude training data. Reddit alleged that Anthropic used Reddit user data without consent. ¹⁷

In January 2024, the Federal Trade Commission (FTC) issued information requests to Anthropic and its corporate financial backers Alphabet, Amazon, and Microsoft as well as rival AI company OpenAI over concerns of tech monopolization. ¹⁸