Anthropic negotiates purchase of AI startup for US$6 billion, agency says

Why are so many people talking about Claude? Meet the most dangerous AI in the world Anthropic, the company responsible for the Claude chatbot, is negotiating the purchase of the artificial intelligence startup Decart AI in a deal that could reach approximately US$6 billion (R$31 billion), the information is from the Bloomberg agency. Decart is an artificial intelligence startup founded in Israel that develops infrastructure technologies to make AI models faster and more efficient. The company also creates its own artificial intelligence systems and has the giant Nvidia among its investors. ??Do you have any reporting suggestions? Send to g1 In May, Decart raised US$300 million in a round led by Radical Ventures. Nvidia participated in the fundraising as a new investor, alongside companies such as Adobe, Amazon and Toyota Ventures. With the round, the startup is now valued at around US$4 billion. The potential acquisition comes as Anthropic seeks to expand its processing capacity to meet increased demand for Claude and prepares for a possible IPO. 'Circular investment': how money moves between AI giants and fuels fears of a new bubble According to information released by the agency, negotiations are still at an early stage. If the deal is completed, the Decart team should be incorporated into the Anthropic area responsible for improving the performance and efficiency of artificial intelligence models. The Claude chatbot, from Anthropic, is among the most popular, rivaling ChatGPT and Gemini Reuters via BBC Among Decart's products is Lucy, a model capable of editing videos in real time. The startup also developed Oasis, which creates virtual environments to train and test robots and autonomous driving systems. Claude, ChatGPT's rival, will reveal who created text with AI; see how it will work Anthropic has been increasing investments in infrastructure to accompany the expansion of its services. Last week, the company announced that it is hiring hardware and software engineers to help develop faster, more efficient custom chips and AI models. Meet Anthropic Anthropic, creator of Claude, was founded in 2021 by Dario Amodei (current CEO of the company) and other executives who left OpenAI defending a more cautious approach to the development of AI. Claude only reached the public in March 2023, around five months after the launch of ChatGPT. Even though it entered the dispute later and was still less known to the general public, the chatbot helped drive Anthropic's impressive turnaround. Today, it is the most valuable AI startup in the world. Valued at around US$965 billion (R$4.87 trillion), the company already surpasses OpenAI in market value. In early June, Anthropic confidentially filed an IPO in the United States. With a valuation close to US$1 trillion, the company could become part of the most select group of companies listed on the American market. Company has the most dangerous AI model in the world Claude's models have proven to be increasingly powerful. "They tend to get things right when they launch. You can see great care in the development of products, which generally reach the market with very competitive performance", says Izabela Anholett. At the same time, experts warn that Anthropic's most advanced systems also increase the risks associated with artificial intelligence. In early 2026, the company announced Claude Mythos, considered by experts to be the most dangerous AI model in the world today. The system is described as extremely fast and capable of finding security holes that would go unnoticed by humans. “If Mythos ends up in the wrong hands, companies could be paralyzed overnight,” says Izabela. According to her, the risk lies in the model's ability to identify and exploit vulnerabilities in digital systems. "Mythos Preview [test version] has already found thousands of high-severity vulnerabilities, including some in all major operating systems and web browsers," Anthropic said. "Due to its high power and the risks it poses to national and general security, it was not released to the common public. Instead, Anthropic created a consortium of companies and government agencies with access to the tool", explains Diogo Cortiz. The group's members include Amazon, Apple, Microsoft, Google, Nvidia, Broadcom and Mozilla. Cortiz recalls that an example of the model's effectiveness came from a Mozilla report. By using Mythos to analyze the Firefox browser, the organization identified a large number of previously unknown vulnerabilities and flaws. "As part of our ongoing collaboration with Anthropic, we had the opportunity to use an early version of the Claude Mythos Preview in Firefox. Version 150 of the browser, released this week, brings fixes for 271 vulnerabilities identified during this initial assessment," Mozilla said in a report on the topic published in April. For Cortiz, although the marketing of AI companies can sometimes exaggerate the capabilities of their products, cases like this show that Mythos' ability to find security holes is real — and help explain why its access is so restricted.