Iren shares climbed 5% in premarket trading after reports that the company signed a multi-year cloud computing deal with Anthropic. The agreement gives Anthropic access to high-performance AI infrastructure, potentially at lower cost and with more flexibility. Anthropic currently relies on Amazon and Google for computing capacity, so this partnership with Iren adds another option. Financial terms of the deal were not disclosed.
Microsoft CEO Satya Nadella pushed for stronger AI safeguards, arguing that companies should treat frontier AI models as potential insider threats. He called for tamper-resistant logs, third-party testing, and controls that operate independently of the model itself. Nadella also urged organizations to disclose failures and share details publicly so others can improve their defenses. Microsoft has already taken steps in this direction.
micro1 announced a $1 billion plan over 12 months to buy and license business data for AI training through its Company Data Partnerships program. The initiative uses de-identified operational data to build reinforcement learning environments, with financing from Citi and Hercules Capital. The goal is to help AI models learn complex real-world business tasks.
Banks are shifting how they measure AI success, focusing on financial returns rather than just hours saved. Bank of America CEO Brian Moynihan reported 130 to 140 AI projects costing $400 million that generate $800 million in benefits, and the bank plans to double its AI budget next year. PYMNTS Intelligence found that 90% to 100% of large U.S. enterprises see returns on AI spending.
The City of Monroe Common Council approved a policy limiting AI use on city networks, passing the measure on October 5 with no opposition. The policy covers tools like ChatGPT, Claude, and Google Gemini, but MS Co-Pilot is now the official AI tool for city employees because it keeps data encrypted and does not train other models. Mayor Tom Miller said network security was the main goal.
PayJoy is acquiring key assets of Finia AI to strengthen its AI underwriting and financial product tools. Finia AI built systems to assess creditworthiness for micro-businesses with little formal credit history. PayJoy serves over 20 million customers across nine countries, and the deal aims to bring Finia's technology to its global platform.
A lawsuit claims McDonald's AI pricing tools lead to higher fast-food prices. The system uses machine learning to analyze millions of daily transactions across nearly 14,000 restaurants and sets optimal prices for each menu item at each location. The lawsuit says the system widened price gaps between stores, while McDonald's called the lawsuit filled with inaccuracies.
Russia's Ministry of Defense created a service giving drone developers access to video data from real drone operations to train military AI models. More than 40 organizations are connected to the platform, and the ministry also built a standard classifier to help developers work with the data consistently.
A new study tested AI models for sorting patent claims without supervision, training hyperbolic and Euclidean autoencoders on 45,000 US patents. The hyperbolic model found cross-category prior art 52% more often than the Euclidean model. Researchers also analyzed 50,000 patents across 10 technology domains and found that the breadth of a patent category affects model performance.
Premier hired Florian Quarré as Chief AI and Analytics Officer and Matthew Zubiller as product leader. Quarré will guide enterprise AI, data, analytics, and technology strategy and lead a new AI & Analytics Center of Excellence as part of Premier's effort to modernize its healthcare technology portfolio.
Key Takeaways
- Iren shares rose 5% after reports of a multi-year AI cloud deal with Anthropic, which could give Anthropic more GPU flexibility at lower cost.
- Anthropic currently relies on Amazon and Google for computing capacity, making the Iren deal a new infrastructure option.
- Microsoft CEO Satya Nadella called for AI safeguards that assume models are compromised, including tamper-resistant logs and third-party testing.
- micro1 pledged $1 billion over 12 months to buy and license business data for AI training, financed by Citi and Hercules Capital.
- Bank of America reported $800 million in benefits from 130 to 140 AI projects costing $400 million, and plans to double its AI budget.
- 90% to 100% of large U.S. enterprises see returns on AI spending, according to PYMNTS Intelligence.
- The City of Monroe banned most AI tools on its network but designated MS Co-Pilot as the official AI tool due to encryption and no model training.
- PayJoy is acquiring Finia AI assets to improve underwriting for micro-businesses with limited credit history, serving over 20 million customers across nine countries.
- A lawsuit claims McDonald's AI pricing tools across nearly 14,000 restaurants lead to higher fast-food prices, which McDonald's disputes.
- A study found hyperbolic AI models outperformed Euclidean models by 52% in finding cross-category prior art across 45,000 patents.
Iren shares jump 5% on reported Anthropic AI cloud deal
Iren shares rose 5% in premarket trading after a report that the company signed a multi-year AI cloud deal with Anthropic. The deal shows Iren can run high-performance AI computing centers. It also points to strong demand for AI infrastructure from third-party providers. The deal could help Anthropic lock in GPU clusters at lower cost. Financial terms were not disclosed.
IREN stock rises on report of multiyear Anthropic cloud deal
IREN Limited stock rose after reports of a multiyear cloud deal with Anthropic. The company had previously disclosed a contract with a major AI lab but kept the customer anonymous. Anthropic's growing need for computing resources is driving demand for data center capacity. Companies with bitcoin-mining backgrounds like IREN are expanding into this space. The deal's financial terms and computing commitments remain undisclosed.
Iren shares rise 5% on reported multi-year AI cloud deal with Anthropic
Iren shares climbed 5% in premarket trading after reports of a multi-year agreement to provide AI cloud computing to Anthropic. The deal would give Anthropic access to high-performance computing infrastructure as it expands. Anthropic currently relies on Amazon and Google for computing capacity. The deal could let Anthropic secure more GPU clusters with greater flexibility and lower costs. Iren has its own data centers and power supplies. Financial terms and computing commitments were not disclosed.
Nadella urges AI safeguards that assume a model is compromised
Microsoft CEO Satya Nadella published an article calling for AI safeguards that assume a model is compromised. He argues that AI models cannot always be traced like traditional software. He wants permissions and activity logs kept outside the AI's reach with independent validation. Nadella rejects unreadable AI reasoning as an excuse to hide problems. He also calls for ongoing testing and incident disclosure. Buyers should ask vendors to prove authorized operators can stop active workflows and preserve untampered logs.
Microsoft CEO Satya Nadella called for an emergency brake on advanced AI models
Microsoft CEO Satya Nadella called for an emergency brake on advanced AI models. He urged companies to treat frontier AI models as potential insider threats. Nadella said organizations must assume a model is compromised and contain it from the start. He called for tamper-resistant logs, third-party testing, and controls outside the model itself. He also urged companies to disclose failures and share details so others can improve defenses. Microsoft had already taken steps in this direction.
micro1 pledges $1 billion for enterprise AI training data
micro1 announced a $1 billion plan over 12 months to buy and license business data for AI training. The effort runs through its Company Data Partnerships program. It will use de-identified operational data to build reinforcement learning environments. The goal is to help AI models learn complex real-world business tasks. Citi and Hercules Capital are financing the initiative.
Banks shift from AI usage metrics to financial returns
Banks are now measuring AI success by money earned, not just hours saved. Bank of America CEO Brian Moynihan reported 130 to 140 AI projects costing $400 million that generate $800 million in benefits. The bank plans to double its AI budget next year. Société Générale and Commerzbank also reported major expected AI savings. PYMNTS Intelligence found 90% to 100% of large U.S. enterprises see returns on AI spending.
Study tests AI models for classifying patent claims
A new study tested AI models for sorting patent claims without supervision. Researchers trained hyperbolic and Euclidean autoencoders on 45,000 US patents. The hyperbolic model found cross-category prior art 52% more often than the Euclidean model. The study also analyzed 50,000 patents across 10 technology domains. It found that how broad a patent category is affects model performance.
Lawsuit says AI pricing raises fast-food prices
A lawsuit claims McDonald's AI pricing tools lead to higher fast-food prices. The system uses machine learning to analyze millions of daily transactions across nearly 14,000 restaurants. It sets optimal prices for each menu item at each location. The lawsuit says the system widened price gaps between stores. McDonald's called the lawsuit filled with inaccuracies.
Monroe city approves rules for employee AI use
The City of Monroe Common Council approved a policy limiting AI use on city networks. The measure passed on October 5 with no opposition. Mayor Tom Miller said the main goal was network security. The policy covers tools like ChatGPT, Claude, and Google Gemini. MS Co-Pilot is now the official AI tool for city employees because it keeps data encrypted and does not train other models.
Premier hires new AI and product leaders
Premier has hired Florian Quarré as Chief AI and Analytics Officer and Matthew Zubiller as product leader. Quarré will guide enterprise AI, data, analytics, and technology strategy. He will also lead a new AI & Analytics Center of Excellence. The moves are part of Premier's effort to modernize its healthcare technology portfolio.
Russia shares drone video data for AI training
Russia's Ministry of Defense has created a service giving drone developers access to video data from real drone operations. The data will help train military AI models for use in unmanned systems. More than 40 organizations are now connected to the platform. The ministry also built a standard classifier to help developers work with the data in a consistent way.
PayJoy acquires Finia AI assets to boost underwriting
PayJoy is buying key assets of Finia AI to strengthen its AI underwriting and financial product tools. Finia AI built systems to assess creditworthiness for micro-businesses with little formal credit history. PayJoy serves over 20 million customers across nine countries. The deal aims to bring Finia's technology to PayJoy's global platform and expand access to financial products.
Sources
- Iren shares jump 5% in premarket trade on report of Anthropic AI cloud deal
- IREN Stock Rises on Report of Multiyear Anthropic Cloud Deal
- Iren Shares Rise 5% Premarket Following Reported Multi-Year AI Cloud Deal With Anthropic
- Nadella urges AI safeguards that assume a model is compromised
- Microsoft CEO Satya Nadella called for an 'emergency brake' on advanced AI models
- micro1 Commits $1 Billion To License Enterprise AI Training Data
- Banks Stop Counting AI Hours and Start Counting Returns
- Benchmarking Embedding Geometries and Clustering Algorithms for Unsupervised Patent Claim Classification
- Lawsuit claims AI helps drive higher fast-food prices
- City adopts AI policy for employees
- Premier Names Florian Quarré AI Chief and Matthew Zubiller Product Leader
- Russia Gives Drone Developers Access to a Large Video Dataset for AI Training
- PayJoy to Acquire Strategic Assets of Finia AI, Advancing AI Underwriting and Financial Product Innovation
Comments
Please log in to post a comment.