Anthropic has signed a significant deal with SpaceX to access 300 megawatts of computing power from the Colossus One data center, gaining over 220,000 NVIDIA GPUs within a month. This partnership allows Anthropic to double Claude Code's five-hour rate limits for Pro, Max, Team, and Enterprise plans, while also removing peak hours limit reductions on Claude Code for Pro and Max accounts. The company has raised API rate limits for Claude Opus models and expressed interest in developing orbital AI compute capacity with SpaceX. This deal comes after SpaceX acquired xAI and is now selling GPU access.
AI companies including Anthropic and OpenAI are moving beyond chatbots to handle core banking operations like treasury, compliance, underwriting, and fraud. Banks are increasingly dependent on outside partners as AI deployment timelines shorten, and regulators are treating AI concentration risk and cybersecurity as systemic banking issues. AI vendors are embedding themselves inside banking infrastructure rather than just selling to banks.
Global investment in AI reached nearly $1.6 trillion between 2013 and 2024, but the technology is expected to reduce labor costs by 25% to 40% while displacing millions of workers. Studies estimate 6% to 7% of U.S. jobs, about 10.4 million, could be lost by 2030, with white-collar workers and college graduates hit hardest. These job losses are expected to be permanent, unlike in past recessions.
AI capabilities have doubled since past elections, making it easier for foreign actors to interfere through deepfakes and misinformation. AI agents can automate tasks like creating fake content and launching cyber attacks, though training is available to help election offices use AI for their benefit and prepare for threats. The UAE has launched a national AI security lab in Abu Dhabi to test AI models and applications before government and private sector use, with systems meeting standards receiving a national certification mark.
Anthropic introduced a feature called dreaming for Claude Managed Agents at its Code with Claude conference, allowing agents to review recent events and store important information in memory for future tasks. This feature is currently in research preview and limited to Managed Agents on the Claude Platform. A Stanford GSB class is experimenting with building AI agents that can represent people's preferences in a simulated legislature, though the agents had trouble sticking to the script and understanding trade-offs between issues.
A bill in Harrisburg addresses patient awareness and reducing bias in AI technology used in healthcare, aiming to ensure patients know when AI is being used in their care and prevent unfair or biased decisions. Remzi Arpaci-Dusseau, first dean of the University of Wisconsin Madison's new College of Computing and Artificial Intelligence opening July 1, said the college will prioritize ethics and the public good over industry interests. Canada's Consumer Product Safety Act does not cover digital products like AI models, and Air Canada was held responsible when its AI chatbot gave false information in 2024.
Key Takeaways
- Anthropic signed a deal with SpaceX for 300 megawatts of computing power from the Colossus One data center, gaining over 220,000 NVIDIA GPUs within a month.
- Anthropic doubled Claude Code's five-hour rate limits for Pro, Max, Team, and Enterprise plans and removed peak hours limit reductions for Pro and Max accounts.
- Anthropic raised API rate limits for Claude Opus models and expressed interest in developing orbital AI compute capacity with SpaceX.
- AI companies including Anthropic and OpenAI are moving beyond chatbots to handle banking operations like treasury, compliance, underwriting, and fraud.
- Global AI investment reached nearly $1.6 trillion between 2013 and 2024, with studies estimating 6% to 7% of U.S. jobs (about 10.4 million) could be lost by 2030.
- AI capabilities have doubled since past elections, making foreign interference through deepfakes and misinformation easier and more powerful.
- The UAE launched a national AI security lab in Abu Dhabi to test AI models and applications, with systems meeting standards receiving a national certification mark.
- Anthropic introduced a dreaming feature for Claude Managed Agents, allowing agents to review events and store information in memory for future tasks.
- A bill in Harrisburg addresses patient awareness and reducing bias in AI technology used in healthcare.
- Canada's Consumer Product Safety Act does not cover digital products like AI models, and Air Canada was held responsible when its AI chatbot gave false information in 2024.
Anthropic signs deal with SpaceX for more AI computing power
Anthropic has signed a new deal with SpaceX to use 300 megawatts of computing power from the Colossus One data center. This deal gives Anthropic access to over 220,000 NVIDIA GPUs within a month. The extra computing power will help improve Claude Code for developers and boost capacity for Claude Pro and Max subscribers. SpaceX is selling GPU access after acquiring xAI.
Anthropic raises Claude usage limits and partners with SpaceX
Anthropic announced a partnership with SpaceX to use all the computing power at the Colossus 1 data center, adding over 300 megawatts of capacity. This deal allows Anthropic to double Claude Code's five-hour rate limits for Pro, Max, Team, and Enterprise plans. The company also removed peak hours limit reductions on Claude Code for Pro and Max accounts. Anthropic raised API rate limits for Claude Opus models and expressed interest in developing orbital AI compute capacity with SpaceX.
AI can both help and harm election officials
AI capabilities have doubled since past elections, making it easier for foreign actors to interfere. Risks like deepfakes and misinformation are not new but are now more powerful and widespread. AI agents can automate tasks like creating fake content and launching cyber attacks. However, training is available to help election offices use AI for their benefit and prepare for threats.
UAE opens national AI security lab for certification and cyber resilience
The UAE Cyber Security Council, Cisco, and Open Innovation AI launched a national AI security lab in Abu Dhabi. The lab will test AI models, autonomous agents, and applications before they are used by government and private sectors. Systems that meet standards will receive a national certification mark. The lab uses Cisco AI-ready infrastructure with NVIDIA GPUs and Open Innovation AI platforms.
AI companies compete to become banking's technology backbone
Anthropic, OpenAI, and other AI firms are moving beyond chatbots to handle treasury, compliance, underwriting, and fraud operations in banking. Banks are depending on outside partners because AI deployment timelines are shortening. Regulators are treating AI concentration risk and cybersecurity as banking system issues. AI vendors are embedding themselves inside banking infrastructure rather than just selling to banks.
AI will cause massive job losses and hurt the economy
Global investment in AI reached nearly $1.6 trillion between 2013 and 2024. AI is expected to reduce labor costs by 25% to 40% but will displace millions of workers. Studies estimate 6% to 7% of U.S. jobs, or about 10.4 million, could be lost by 2030. White-collar workers and college graduates will be hit hard, and these job losses will be permanent unlike in past recessions.
AI products are defective and dangerous but we keep using them
Canada's Consumer Product Safety Act does not cover digital products like AI models. AI systems often fail and make mistakes, but companies warn users in their terms of use. Air Canada was held responsible when its AI chatbot gave false information in 2024. The author argues that AI is making things worse and that companies should be held accountable for defective products.
Bill looks at patient awareness and bias in AI healthcare technology
A bill in Harrisburg addresses patient awareness and reducing bias in AI technology used in healthcare. The legislation aims to ensure patients know when AI is being used in their care. It also seeks to prevent AI systems from making unfair or biased decisions. The bill focuses on transparency and fairness in medical AI applications.
UW Madison AI college dean says they are not AI cheerleaders
Remzi Arpaci-Dusseau is the first dean of the University of Wisconsin Madison's new College of Computing and Artificial Intelligence, opening July 1. He said the college will prioritize ethics and the public good over industry interests. The college cannot compete with Silicon Valley salaries but will attract people who want to teach and research. Arpaci-Dusseau wants to bring light and clarity to AI issues and help communities make smart decisions about data centers.
Stanford class experiments with building AI agents to govern
A Stanford GSB class is experimenting with building AI agents that can represent people's preferences in a simulated legislature. Students trained personal AI agents by talking to them about their voting philosophy. The agents voted on shareholder proposals and showed some ability to learn preferences. However, the agents had trouble sticking to the script and understanding trade offs between issues.
Anthropic's Claude Managed Agents can now dream and remember
Anthropic introduced a feature called dreaming for Claude Managed Agents at its Code with Claude conference. Dreaming lets agents review recent events and store important information in memory for future tasks. This feature is currently in research preview and limited to Managed Agents on the Claude Platform. Dreaming helps agents improve their performance and decision making by reflecting on their experiences.
Sources
- Anthropic taps Elon Musk's SpaceX for more AI compute power
- Higher usage limits for Claude and a compute deal with SpaceX
- How AI Can Help (and Hurt) Election Officials
- UAE launches national AI security lab for certification and cyber resilience
- The Battle to Own Banking’s AI Backbone
- The Downside of Artificial Intelligence
- AI products are defective and dangerous. Why are we using them at all?
- State of Artificial Intelligence: Transparency in Healthcare
- Dean of UW’s College of Computing and AI: ‘We’re not here to be AI cheerleaders’
- Training AI To Govern For Us: How This Stanford GSB Class Experiments With Building AI Agents
- Anthropic's Claude Managed Agents can now "dream," sort of
Comments
Please log in to post a comment.