Anthropic has entered a long-term partnership with cloud infrastructure provider CoreWeave to significantly enhance its processing capabilities for developing and deploying its Claude AI models. The deal grants Anthropic access to a vast array of Nvidia GPUs housed in facilities across the United States, addressing the company's urgent need to scale infrastructure amid growing user adoption. This partnership further solidifies CoreWeave's position as a critical backbone for AI projects, as it now supports nine of the world's top ten AI model developers, including OpenAI, Google, and Meta.
This announcement follows Anthropic's recent collaboration with Broadcom and Google to secure approximately 3.5 gigawatts of tensor processing unit (TPU) capacity, expected to ramp up starting next year for large-scale model training in US data centers. Together, these initiatives secure both immediate Nvidia GPU resources and specialized TPUs for future growth, with significant implications for the fintech sector. Enhanced computing power could accelerate Claude's evolution into more sophisticated tools for real-time fraud detection, dynamic risk assessment, and automated regulatory compliance.
Concurrently, the UK's AI Safety Institute released an evaluation of Anthropic's Claude Mythos Preview, a powerful next-generation model. The institute found that the AI can autonomously execute sophisticated cyber attacks, becoming the first model to complete "The Last Ones" (TLO)—a 32-step corporate network attack simulation that typically takes humans 20 hours. In controlled evaluations, Mythos Preview succeeded in three out of 10 attempts, averaging 22 of 32 steps completed.
The model demonstrated an unprecedented 73% success rate on expert-level capture-the-flag tasks, challenges no AI model could complete before April 2025. When given network access, it could autonomously discover and exploit vulnerabilities. This advancement marks a dramatic escalation from just two years ago, prompting warnings from U.S. Treasury Secretary Scott Bessent and Federal Reserve Chair Jerome Powell to bank executives about the looming security threat.
For the cryptocurrency ecosystem, where smart contract vulnerabilities and exchange hacks already cost billions annually, AI-powered attacks represent a new category of potential security threat. DeFi protocols, with their complex interconnected systems, may face particular exposure to automated exploitation attempts that can analyze and attack multiple vectors simultaneously.
