google launches meta while amazon expands its platform

Key Takeaways

• Google is considering natural gas to power its AI data center in North Texas, potentially emitting 4.5 million tons of CO2 annually, reflecting a shift from its clean energy goals.
• Other major tech companies, including Meta, Amazon, and Microsoft, are also exploring natural gas for their data centers to meet high AI energy demands.
• TAG released the "Enterprise AI Security Handbook," a free guide offering practical steps for businesses to secure their AI systems and manage related risks.
• The National Institutes of Health (NIH) is investing $30.7 million to expand the USC-led AI4AD initiative, using AI to classify Alzheimer's subtypes and identify treatment targets.
• IBM launched Granite 4.0 3B Vision, a new vision-language model designed for accurate structured data extraction from enterprise documents, such as converting charts to code.
• Federal AI adoption faces significant challenges due to incomplete, inaccurate, and fragmented data across agencies, hindering the effectiveness and scalability of AI models.
• AI product teams are adopting active, in-product feedback mechanisms to keep pace with rapidly updating AI products, centralizing user input and building trust.
• The European Defence Fund 2026 is investing €16 million in AI-driven simulations to train military leaders for complex, multi-domain operations and human-machine teaming.
• The shutdown of OpenAI's Sora highlights the volatility of AI vendor relationships, urging companies like Disney to develop internal AI strategies and leverage their own data advantage.
• A Darden MBA case competition explored integrating faith and belief into business strategies concerning AI deployment, particularly after an AI agent's inappropriate response led to a lost contract.

Google may use natural gas for AI power needs

Google is considering using more natural gas to power its growing AI operations, a shift from its previous focus on clean energy. The company is partnering with Crusoe Energy on a data center campus in North Texas called 'Goodnight,' which will be powered by a large natural gas plant and a wind farm. While Google has not yet finalized a contract for the gas plant's power, this move highlights the tension between the rapid demand for AI computing power and climate goals. Experts note that natural gas is currently a scalable and available energy source, though it still emits greenhouse gases. This approach reflects a broader trend among tech companies seeking to secure energy for AI while navigating environmental concerns.

Google's Texas AI data center to use gas plant

Google is partnering with Crusoe Energy Systems to build a natural gas power plant for its 'Goodnight' data center campus in Armstrong County, Texas. This plant is expected to emit up to 4.5 million tons of carbon dioxide annually, comparable to the emissions of San Francisco. This development marks a significant shift for Google, which has previously emphasized its commitment to clean energy and a goal of 100% carbon-free operations by 2030. While Google states it does not have a contract for the plant yet, this move aligns with other tech giants like Meta, Amazon, and Microsoft also turning to natural gas for their data centers. The company frames this as part of its 'climate moonshots' and a complex challenge in achieving its goals.

Google's Texas data center to run on gas power

A new Google data center campus in Armstrong County, Texas, named 'Goodnight,' will partially run on power from private natural gas turbines. These turbines are expected to emit over 4.5 million tons of greenhouse gases annually, significantly more than average plants. Google stated it does not have a contract for this gas power, but the project involves both natural gas and wind power. This move by Google, a company known for its clean energy initiatives, suggests a potential shift in strategy to meet the high energy demands of AI. Other tech companies are also exploring natural gas as a power source for their data centers, citing its availability and scalability.

TAG releases AI security guide for businesses

TAG has released a new handbook titled 'Enterprise AI Security Handbook' to help organizations secure their AI systems. The guide offers practical, experience-based advice for security leaders, moving beyond theory to actionable steps. It covers defining AI systems, creating a roadmap for AI security, developing policies aligned with existing frameworks, and prioritizing AI use cases based on risk. The handbook also addresses managing AI-related identities and navigating the AI security vendor market. This free resource aims to provide a clear path for enterprises adopting AI to ensure safe and operational implementation.

TAG offers free AI security handbook for businesses

TAG has launched the 'Enterprise AI Security Handbook,' a free guide designed to help businesses secure their AI adoption. The handbook provides a structured, practical approach for enterprise security leaders, covering AI system definitions, security roadmaps, policy development, and risk assessment. It also offers insights into managing AI identities and understanding the vendor landscape. The guide aims to move beyond theoretical concepts to offer actionable advice for integrating AI securely into existing operations. Experts contributed to the final chapter, discussing the future of AI security.

USC leads AI effort to understand Alzheimer's with $30.7M NIH funding

The National Institutes of Health (NIH) is investing $30.7 million to expand the USC-led Artificial Intelligence for Alzheimer's Disease (AI4AD) initiative. The new phase, AI4AD2, will use AI, brain imaging, genetics, and multi-omics data to better classify Alzheimer's subtypes, predict disease progression, and find new treatment targets. This project unites researchers from 10 institutions to analyze large datasets, including genetic and brain scan information. The goal is to develop 'genomic language models' to identify genetic factors linked to Alzheimer's and improve diagnosis and treatment, especially for diverse global populations.

Darden MBA students compete in faith and AI case challenge

Three University of Virginia Darden School of Business MBA students participated in the Fourth Annual Faith and Belief at Work Case Competition at Brigham Young University. The competition challenged students to address real-world business problems involving faith, global strategy, and technology, including the deployment of artificial intelligence. The case focused on ServiceNow, an AI software company, and a competitor that lost a major contract due to an AI agent's inappropriate response. The Darden team, mentored by Professor Jared Harris, explored how businesses can better integrate faith and belief into their strategies, especially concerning AI development and deployment in diverse global markets.

IBM launches Granite 4.0 3B Vision for document data extraction

IBM has released Granite 4.0 3B Vision, a new vision-language model (VLM) designed for extracting data from enterprise documents. This model focuses on accuracy for structured data, like converting charts to code or tables to HTML, rather than general image descriptions. It uses a modular approach with a LoRA adapter on top of the Granite 4.0 Micro base model. The system employs a vision encoder and patch tiling to maintain detail in complex document layouts. Trained on specialized datasets like ChartNet, it aims to improve chart and table extraction, showing strong performance on benchmarks like VAREX.

Federal AI adoption hindered by data challenges

The White House's push for federal AI adoption faces a significant hurdle: data readiness. Agencies struggle with incomplete, inaccurate, and fragmented data spread across various environments, which limits the effectiveness and scalability of AI models. This 'garbage in, garbage out' problem can lead to flawed decisions and hinder mission success. Addressing data fragmentation, improving data hygiene through auditing and governance, and treating data as a strategic asset are crucial steps for agencies to overcome these challenges. Without robust data readiness, the federal government risks stalling its AI initiatives before they can deliver meaningful impact.

AI product teams rethink customer feedback in 2026

AI product teams in 2026 are transforming how they gather customer feedback to keep pace with rapidly updating AI products. Traditional methods are insufficient for AI products that change weekly and face intense competition. Teams are now building active feedback mechanisms directly into the product experience, allowing users to report issues and request features in context. Tools like Frill are centralizing feedback, enabling users to submit ideas, vote on them, and track progress via a public roadmap. This approach builds user trust and retention by making the feedback-to-shipping process transparent.

AI simulations to train future military leaders

The European Defence Fund (EDF) 2026 is investing €16 million in AI-driven simulation and training for modern defense. Traditional training methods struggle with the complexity of hybrid threats and multi-domain operations. AI-powered simulations can generate realistic, dynamic scenarios, allowing military leaders to practice decision-making in a risk-free environment. Use cases include assessing multi-domain operations performance, exploring human-machine teaming, assisting in course of action development, and generating synthetic data for AI training. This initiative aims to create more agile, adaptable, and trustworthy military capabilities for the future.

Lessons learned from AI vendor failures like Sora shutdown

The sudden shutdown of OpenAI's Sora video generator highlights the volatility of AI vendor relationships and the importance of building internal AI strategies. Companies like Disney, which had a major deal with Sora, faced significant disruption. Experts advise organizations not to outsource their AI strategy entirely, as products can be launched and canceled rapidly. Building an internal 'data advantage' is crucial, as demonstrated by Google's extensive real-world data from Waymo and Street View, which provides a competitive edge. The article also points to medical publishers missing opportunities by not leveraging their data effectively, contrasting with startups like OpenEvidence that successfully monetize data streams.

Sources

• Google eyes natural gas as AI power demand outpaces clean energy
• Google teams up with gas plant for AI datacenter in sharp turn from climate goals
• A New Google-Funded Data Center Will Be Powered by a Massive Gas Plant
• TAG Releases AI Security Handbook to Guide Organizations in Securing AI
• TAG Releases AI Security Handbook to Guide Organizations in Securing AI
• NIH investment totaling 30.7M will expand USC-led AI effort to decode Alzheimer’s disease
• When AI Meets Faith: Lessons From Darden’s Participation in an MBA Case Competition on Faith and Belief at Work
• IBM Releases Granite 4.0 3B Vision: A New Vision Language Model for Enterprise Grade Document Data Extraction
• The data challenge impacting federal AI adoption
• How AI Product Teams Are Rethinking Customer Feedback in 2026
• AI-Driven Defence Simulations & Military Training Future
• When AI Vendors Fail: Lessons From The Sora Shutdown