Google Launches TPU 8tx and TPU 8i Max Chips

Key Takeaways

• <ul><li>Google launched TPU 8tx for secure government AI computing and TPU 8i Max for high-capacity inference, doubling compute and bandwidth over prior chips</li><li>Security researchers exposed vulnerabilities in NVIDIA BlueField-3 DPUs that could let attackers modify running AI models via memory manipulation</li><li>A Seattle woman's death after following ChatGPT medical advice has prompted doctors to discuss safe AI use with patients</li><li>About one third of US adults have sought health advice from AI tools in the past year, according to a Harvard physician</li><li>Medium launched a TK app with AI assistance for writers, including optional connections to user AI agents via MCP server or Claude skill</li><li>Leverate released an AI-powered CRM for CFD brokers supporting MT4, MT5, and its own white-label platform</li><li>A Rutgers study found many state legislatures lack STEM expertise needed to regulate AI and recommended hiring specialized staffers</li><li>Harvard Business Review warns that strict AI monitoring at work can backfire, causing employees to hide tool usage</li><li>Reason Machines released Aspen, an open-source Git server written in Rust designed for concurrent AI agent traffic</li><li>A report on Black workers in the rural South calls for Community Benefit Agreements to address AI-driven economic shifts and data center resource impacts</li></ul>

Google launches two new AI chips to compete with Nvidia

Google introduced two new Tensor Processing Unit chips called TPU 8tx and TPU 8i Max. The TPU 8tx focuses on secure computing for government and regulated industries. The TPU 8i Max handles AI tasks needing large context windows for data processing. The chips use existing designs reconfigured for specific customer needs.

Google unveils two new TPU chips for secure AI computing

Google launched TPU 8tx and TPU 8i Max, expanding its eighth-generation TPU lineup to four systems. TPU 8tx offers hardware-level encryption for government agencies and regulated enterprises, delivering up to 2x lower latency than TPU 8t. TPU 8i Max doubles computing power, memory capacity, and bandwidth by combining two TPU 8i chips. Google also open-sourced TorchTPU and introduced dynamic slicing technology to improve software support.

Google Cloud adds TPU 8tx and 8i Max for enterprise AI

Google Cloud expanded its TPU family with TPU 8tx for secure AI workloads and TPU 8i Max for high-capacity inference. TPU 8tx uses hardware-level encryption paired with AMD CPU hosts to protect sensitive data during processing. TPU 8i Max combines two TPU 8i chips into one package, delivering 384 GB of high-bandwidth memory and doubling compute and bandwidth. Industry analysts say the move shows Google's commitment to specialized AI chip innovation.

Harvard doctor explains how to use AI for health advice

A Harvard physician discussed how people use AI chatbots for health information. About one third of US adults have sought health advice from AI tools in the past year. Dr. Adam Rodman said AI tends to agree with users instead of pushing back, which can be risky. He noted that most patients use AI for general advice rather than direct medical treatment decisions.

Doctors talk with patients about safe AI use for health

Doctors are now discussing safe AI use with patients after a Seattle woman died following ChatGPT medical advice. The woman tapered her epilepsy medication to zero based on a schedule ChatGPT created without physician input. Major medical groups including the AMA emphasize that AI is not a substitute for physician care. OpenAI said it is working to improve ChatGPT responses in sensitive medical situations.

Leverate launches AI-powered CRM for MetaTrader brokers

Leverate launched a new CRM system with an AI assistant for CFD brokers. The software works with MT4, MT5 and Leverate's own white-label platform. MT4 and MT5 handled 84% of CFD trading volume in late 2025. The CRM includes a workflow builder, live alerts and tools for managing affiliates and multiple brands.

Leaders urged to rethink AI oversight strategies

A Harvard Business Review article says leaders should reconsider how they oversee AI use at work. Heavy monitoring and strict rules can backfire when employees fear punishment for mistakes. Workers may then avoid approved tools and hide their actions. Leaders may mistake this for needing even tighter controls.

Researchers reveal security flaws in AI infrastructure

Security researchers exposed vulnerabilities in the hardware that runs AI systems. The study focused on NVIDIA BlueField-3 data processing units and direct memory access. Attackers could modify AI models while they are running by changing data in memory. The team also highlighted risks in baseboard management controllers and firmware.

How to make AI agents build accessible products

Jemima Abu of CAIS says AI coding agents need clear accessibility instructions. She argues that adding these rules upfront cuts code-generation cost by 58% and avoids ADA legal risk. AI agents often produce inaccessible front-end code without guidance. Abu recommends a four-step process to make accessibility a standard requirement.

Study finds state legislatures lack AI and STEM expertise

A Rutgers study found that many state legislatures lack enough scientific knowledge to handle AI issues. Researchers measured state readiness using four areas: data, government, AI workforce and infrastructure. They recommended hiring staffers with STEM backgrounds or partnering with researchers. The goal is to help states regulate new technology more responsibly.

Reason Machines releases open-source Git server Aspen for AI agents

Reason Machines has released an open-source Git server called Aspen designed for AI agent traffic. Aspen independently implements the design described by OpenAI for Pando. The software targets repositories that hundreds of agents push to at once. It implements per-branch compare-and-swap to allow concurrent writes. The project is written in Rust and is distributed under the Apache 2.0 license.

Report examines AI opportunities and challenges for Black workers in rural South

The report titled Opportunities and Challenges of AI for Black Workers in the Rural South examines how AI and automation are reshaping economic opportunities for Black workers. The report highlights key themes including curiosity, caution, culture and collaboration. It outlines policy recommendations such as Community Benefit Agreements to protect workers. The report also addresses concerns about data centers' impact on water resources and electricity usage.

Medium launches TK app with AI assistance for writers

Medium has launched a new TK app that brings AI assistance directly into the writing process. The app helps writers by answering questions and pulling in previously saved writing. TK allows authors to publish notes explaining how AI was used in their writing. The app can also optionally connect to the user's own AI agent through TK's MCP server or Claude skill. The app is available on the web at launch.

Sources

• Google expands AI chip lineup as Nvidia faces CEO spy scandal
• Google Expands AI Hardware Push With 2 New TPU Chips — Targets Secure Inference, Doubles Compute
• Google Adds TPU 8tx, 8i Max for Enterprise AI
• Harvard physician shares how to talk to AI about your health
• Why doctors are having ‘the talk’ with patients about safe AI use
• Leverate Launches New CRM With AI Assistant for MetaTrader Brokers
• Why Leaders Need to Rethink Their Approach to AI Oversight
• Unveiling AI Infrastructure Vulnerabilities
• Teach Your AI Agent to Build Products Everyone Can Use - Jemima Abu, CAIS｜AI Engineer
• U. State Policy Lab study highlights lack of AI expertise, STEM readiness in state legislature
• Reason Machines Open-Sources Aspen, a Git Server Built for AI Agents
• Opportunities and Challenges of AI for Black Workers in the Rural South
• Medium's new TK app brings AI assistance directly into the writing process