AI News (07/27) : Nvidia backs massive OpenAI data center as public Claude chats expose private data

The boundaries of AI infrastructure, model efficiency, and data privacy are shifting fast today. As Nvidia weighs unprecedented financial guarantees to back a half-trillion-dollar OpenAI supercluster, Anthropic faces a fresh security headache over exposed Claude chats, and Moonshot AI challenges Western dominance with a highly optimized open-weight release.

1. Anthropic’s Claude Chats Left Searchable on Google

A trove of chat logs and projects created with Anthropic’s Claude—some containing sensitive medical reports, clinical trial results, internal corporate documents, and employee reviews—have been indexed by Google and made searchable to the public. The exposure occurs when users use Claude's "share" feature to distribute links to colleagues. While Anthropic warns that shared links are "public," it does not explicitly alert users that these links can be crawled and indexed by search engines. Anthropic stated that the system is working as intended, noting that links are not discoverable unless shared publicly by users. Why it matters: Generative AI tools remain a major vector for accidental corporate and personal data leaks, highlighting the gap between user privacy expectations and public sharing mechanisms.

2. Nvidia Weighs Backing $500 Billion OpenAI Data Center Project

Nvidia is reportedly in talks to provide an unprecedented $250 billion financing guarantee to help OpenAI secure a massive 10-gigawatt data center in southern Ohio, developed by SoftBank. Nvidia is also exploring an additional $350 billion in financing for OpenAI’s chip purchases. The massive Ohio project, expected to cost over $500 billion overall, would allow OpenAI to build its own independent computing infrastructure and reduce its reliance on legacy cloud providers like Microsoft. Why it matters: This astronomical deal would cement Nvidia's absolute dominance in the AI hardware supply chain while enabling OpenAI to establish infrastructure independence.

3. AI Automation Hits Entry-Level Hiring

A new Gartner survey reveals that AI-driven automation is actively reducing entry-level hiring at nearly one-quarter (25%) of surveyed organizations. As AI systems take over routine, entry-level tasks, companies are scaling back traditional junior roles. Why it matters: As AI systems assume routine junior tasks, the traditional corporate talent pipeline is tightening, forcing new graduates to adapt to a highly competitive job market.

4. Trump Administration to Loosen AI Guardrails

The Trump administration is moving to loosen existing federal AI regulatory guardrails. The shift is designed to reduce compliance burdens on developers and accelerate domestic deployment. Why it matters: Rolling back AI regulations may fast-track commercial innovation in the U.S., but it risks intensifying safety concerns and complicating international efforts to establish unified safety standards.

5. Anthropic Releases Cost-Efficient Claude Opus 5

Anthropic has launched Claude Opus 5, an updated hybrid reasoning model. Priced at the same rate as its predecessor ($5 per million input and $25 per million output tokens), Opus 5 delivers a massive leap in coding, agentic tasks, and professional reasoning. Anthropic claims the model nearly matches the frontier capabilities of its premium Claude Fable 5 model at half the cost per task. Why it matters: AI labs are pivoting from pursuing raw performance at all costs to engineering highly optimized, cost-efficient reasoning models suitable for commercial agents.

6. Nvidia Invests Billions in Safe Superintelligence

Nvidia has made a multibillion-dollar investment in Safe Superintelligence (SSI), the high-profile startup co-founded by former OpenAI Chief Scientist Ilya Sutskever. Why it matters: Nvidia is diversifying its strategic bets beyond cloud providers by directly funding pure-play, frontier-level safety and superintelligence research.

7. BYD to Unveil First Humanoid Robot in August

Chinese electric vehicle giant BYD announced plans to publicly debut its first humanoid robot in early August. BYD executives intend to deploy these robots across the company's vast dealership network to act as sales assistants, demonstrate vehicle features, and engage customers. Why it matters: Humanoid robots are transitioning from back-end manufacturing floors to consumer-facing retail and service roles much faster than industry analysts predicted.

8. Google’s AI Overviews Become the Search Default

A new report from Similarweb shows that Google’s AI Overviews now appear in 43% of all search queries, up from just 15% a year ago. Visits to Google's conversational AI Mode rose to 279 million by May 2026. This shift turns Google into a final destination rather than a gateway, hurting publisher referral traffic. Why it matters: Google's transition to an AI-first answer engine is disrupting the economic foundations of the open web, starving publishers of traffic while locking users into its ecosystem.

9. Nvidia Launches Real-Time Surgical Simulator Cosmos-H-Dreams

Nvidia has introduced Cosmos-H-Dreams, an action-conditioned, real-time generative simulator for surgical robotics. Running on a single RTX PRO 6000 GPU, the model translates live robot kinematics into realistic video streams of deformable tissues and instrument interactions, allowing surgical AI policies to train in a closed loop. Why it matters: Real-time visual world models enable safe, physical-free testing and synthetic data generation for highly complex, high-stakes robotic environments like surgery.

10. Moonshot AI Releases Weights for High-Performance Kimi-K3

China’s Moonshot AI has released the open weights for its 2.8-trillion-parameter Kimi-K3 model. Kimi-K3 closely rivals frontier U.S. models like Claude Fable and GPT-5.6 Sol on benchmarks while being up to ten times cheaper to run due to extreme architectural optimizations, including low-precision math and a fixed-size attention mechanism. Why it matters: High-quality, open-weight models from Chinese competitors are narrowing the capabilities gap with proprietary Western giants while radically lowering the cost of frontier-level intelligence.

To stay up to date on everything going on in AI, check out the tracker at the500feed.com

---

Sources

1. A Whole Bunch of People’s Claude Chats Are Publicly Accessible Online, and There’s Some Wildly Private Stuff in There
2. The AI data center behind ChatGPT could cost more than $500 billion
3. Gartner Survey Finds AI Automation Is Reducing Some Entry Level Hiring at Nearly One-Quarter of Organizations
4. Trump is loosening AI guardrails. What does that mean for you?
5. Anthropic releases ‘more efficient’ Claude Opus 5 AI model
6. Nvidia Makes Multibillion Dollar Investment in Ilya Sutskever’s Safe Superintelligence
7. BYD to unveil first humanoid robot in early August
8. Google’s AI search is rapidly becoming the default, new data shows
9. NVIDIA Cosmos-H-Dreams: Bringing Real-Time Generative Simulation to Surgical Robotics
10. Moonshot AI releases weights for Kimi-K3, firing a shot across the bow of OpenAI and Anthropic — open-weight model performs almost as well as frontier models while being 2-3x easier to run