The AI Impact Summit venue in New Delhi, India, on Nov. 19. As Indian Prime Minister Narendra Modi raised the hands of AI leaders, including Google CEO Sundar Pichai, for a commemorative photo, camera flashes went off everywhere. The cameras focused not on Modi but on two AI titans: OpenAI CEO Sam Altman and Anthropic CEO Dario Amodei. All the participants on the stage raised the hands of those beside them and smiled, but the two were an exception. Refusing to shake hands with each other, they wore awkward expressions and bitter smiles while staring straight ahead.
The scene spread rapidly across social media. Siddharth Bhatia, co-founder of AI startup Pook AI, posted on X (formerly Twitter): "When will artificial general intelligence arrive? Probably the day Dario and Sam join hands." Justin Moore, an investment partner at Andreessen Horowitz, shared the photo, in which the discomfort was evident, with the comment: "When you're forced to do a group project with your boss."
The two CEOs are now irreversible rivals, but Altman and Amodei were originally part of the same family at OpenAI. After working as a senior researcher at Google Brain, Amodei joined OpenAI and led the development of large language models (LLMs) such as GPT-2 and GPT-3, laying the foundation for ChatGPT. Amodei served as vice president of research at OpenAI, while his sister, Daniela Amodei, served as vice president of safety and policy, playing key roles in the company's management.
But when OpenAI sought to attract Microsoft (MS) investment to raise funds and convert from a nonprofit to a for-profit entity, Amodei came into conflict with the company. At the time, deeply concerned with AI ethics and safety, he criticized the Microsoft deal as one that would steer OpenAI in an overly commercial direction.
In the end, the Amodei siblings left the company in 2020 and founded Anthropic with colleagues who departed alongside them. The idea was to build AI under strict control. The company grew into a major competitor, attracting billions of dollars in investment from Google, Salesforce, and Amazon, and threatened OpenAI's ChatGPT by releasing the AI chatbot "Claude."
Amodei's intention to build AI models different from OpenAI's was clearly revealed in a 2023 Fortune interview. Asked by the host why he left OpenAI, he answered: "I wanted to build more trustworthy models" and "I founded Anthropic because of concerns about OpenAI."
"After developing GPT-2 and GPT-3 at OpenAI, I had two very strong beliefs. The first was that as you put more compute into the models, their performance keeps improving, and there is no end to those improvements. The second was that, beyond scaling the models, you need an element called alignment or safety. Increasing compute alone cannot set the value standards for a model. I founded the company based on these ideas." (Fortune interview)
Ahead of this year's initial public offering (IPO), the rivalry between OpenAI and Anthropic has intensified further. Anthropic's valuation of about $380 billion (562.21 trillion won) is chasing OpenAI's corporate value of $840 billion. Previously, OpenAI focused on general users while Anthropic centered on enterprise services for corporate clients, so there was a strong perception that OpenAI's rival was Google, which developed "Gemini." But recently the atmosphere has changed considerably. The perception spread that Claude Cowork and Claude Code are leading AI products threatening the software market, and OpenAI, feeling a sense of crisis, is now focusing on securing corporate clients.
In particular, as OpenAI pushed to introduce advertising into its AI services from last year to improve profitability, the nerve war with Anthropic heated up. Last month, an advertisement during the Super Bowl, the championship game of the National Football League and the largest sports event in the United States, became a decisive blow. Anthropic aired scenes showing an AI giving absurd answers—recommending a dating site to a son trying to improve his relationship with his mother, or suggesting height-boosting insoles in response to a question about how to quickly reap the benefits of exercise—which was a criticism of ChatGPT's advertising. Altman did not hide his displeasure, calling it "deceptive and dishonest."
Recently, the two companies clashed again as their positions diverged over the AI policy of the Donald Trump administration. The U.S. Department of Defense (Department of War)—which this year captured Venezuelan President Nicolás Maduro alive and then launched airstrikes on Iran—pressured Anthropic, saying it would use Claude for all purposes without restriction. But Anthropic refused, stating that AI should not be used for mass surveillance of citizens or fully autonomous weapons. In the end, on Nov. 27, President Trump condemned Anthropic as "radical left" and urged all federal agencies to stop using Anthropic technology. Defense Secretary Pete Hegseth designated Anthropic as a supply-chain risk factor—the first time for a domestic company—blocking companies that use Claude from doing business with the Department of Defense.
The U.S. Department of War signed a usage contract with OpenAI to fill Anthropic's vacancy, but OpenAI ran into a user backlash. Unlike Anthropic, which resisted the government, OpenAI was assessed as having capitulated to the Trump administration, and a phenomenon is occurring in which customers are switching en masse to Anthropic. According to Anthropic, the number of free Claude users has increased by more than 60% since January this year, and the number of paid subscribers has doubled this year. On Nov. 28, Claude overtook OpenAI's ChatGPT to take the No. 1 spot in the free app rankings on the U.S. Apple App Store.
For customers switching from ChatGPT to Claude, Anthropic has been offering a free import feature since the 2nd. It enables users to copy content entered into ChatGPT and continue searching on Claude. Business Insider analyzed that "Claude reached No. 1 on the App Store because ChatGPT users departed while expressing support for Anthropic."
As public opinion turned unfavorable, OpenAI belatedly moved to amend its Department of Defense contract. On the 2nd, Altman said via X (formerly Twitter): "We are working on adding clauses with the Department of Defense to clarify our principles," announcing that he would include language in the contract stating that AI systems should not be intentionally used for domestic surveillance of U.S. residents or nationals. He even showed signs of reflection, saying: "We shouldn't have given hurried notice on Nov. 27. This issue is very complex and required clear communication," and "As a result, I think we looked opportunistic and sloppy."
