Anthropic, the developer of Claude, is in negotiations to acquire Decart AI, an Israeli artificial intelligence startup that was valued at approximately US$ 4 billion. However, the transaction could reach a value of around US$ 6 billion.
This acquisition is part of Anthropic's strategy to strengthen its computational capacity and keep pace with the expansion of Claude. Decart AI creates infrastructure solutions focused on increasing the speed and efficiency of AI systems, in addition to possessing its own models.
Although the conversations are in a preliminary stage, if the deal is finalized, Decart's employees will be integrated into the Anthropic team responsible for the performance and efficiency of its models. This move comes as the company prepares for a possible Initial Public Offering (IPO) in the United States.
Decart raised US$ 300 million in an investment round led by Radical Ventures. Nvidia joined the investors, alongside other companies such as Adobe, Amazon, and Toyota Ventures. After this fundraising, the startup's valuation was estimated at around US$ 4 billion.
Among the technologies developed by the company, Lucy, a system for real-time video editing, and Oasis, used to create virtual environments for training and evaluating robots and autonomous driving technologies, stand out.
The potential purchase aligns with Anthropic's effort to expand its processing resources. The company has been increasing infrastructure spending to meet the demand for its services and, last week, announced the hiring of hardware and software specialists to work on proprietary chips and faster, more efficient AI models.
Anthropic was founded in 2021 by Dario Amodei and other executives who left OpenAI. It launched Claude in March 2023, about five months after the launch of ChatGPT, and has since gained ground in the AI market.
The material indicates that Anthropic is valued at around US$ 965 billion and has already surpassed OpenAI in market value. At the beginning of June, the company also confidentially filed an application to conduct an initial public offering in the United States.
In addition to commercial and infrastructure expansion, Anthropic gained notoriety with Claude Mythos, presented in early 2026. This model was described as having the ability to identify security flaws that might go unnoticed by human analysis.
Experts commented that this advancement in capabilities also raises risks associated with the use of AI systems. Izabela Anholett opined that Anthropic's products generally reach the market with good performance and demonstrate care in their development. However, she warned about the danger if Mythos were used by malicious individuals, due to its aptitude for detecting and exploiting flaws in digital environments.
Due to the inherent risks of the tool, access to Mythos has not been released to the general public. Anthropic created a group composed of companies and government agencies authorized to use the system, including Amazon, Apple, Microsoft, Google, Nvidia, Broadcom, and Mozilla.
Mozilla also serves as a practical example of the model. An initial version of Claude Mythos Preview was used in the analysis of Firefox, aiding in the discovery of vulnerabilities that led to 271 fixes in version 150 of the browser. Diogo Cortiz, another cited expert, believes that such incidents prove the system's ability to find security flaws, but stresses that AI companies may present their products more positively than their actual capabilities.