Blackwell traces in Chinese models, Anthropic’s record IPO, ChatGPT Wallet

November 14, 2026

A former OpenAI researcher says six Chinese AI models carry signatures associated with Nvidia’s Blackwell chips, raising questions about access to export-controlled hardware. Anthropic is drawing record-scale demand for its planned IPO, and OpenAI is bringing payments into ChatGPT. Here’s what changed.

Top Story

🧠 Former OpenAI researcher finds Blackwell traces in six models

Will DePue says six Chinese models released since April 2026 carry a Blackwell signature in their published weights, offering technical evidence of access to export-controlled Nvidia chips. He says senior researchers at OpenAI and Anthropic independently confirmed his findings.

If the claim holds, it could sharpen questions about whether export controls can keep restricted chips out of China’s AI supply chain.

[The Decoder]

Today's stories

📈 Anthropic's IPO draws six times the available orders

Anthropic’s annualized revenue rose to $81 billion by late October, from $71 billion in September. IPO orders are six times the available shares as the Claude developer seeks to raise up to $64.4 billion at a target valuation of roughly $1.4 trillion.

A planned $64.4 billion raise would put investor demand for AI companies on a historic scale.

[The Decoder]

💳 OpenAI gives ChatGPT agents a way to pay

ChatGPT Wallet lets agents pay without seeing users’ card details, while requiring a security check for every purchase and preserving eligible card rewards and dispute protections. Built with American Express, Mastercard and Visa, it will initially be available through dots to U.S. ChatGPT Pro subscribers.

The move brings agent spending into a mainstream chat product, with a security check for each purchase.

[OpenAI News]

🛡️ Check Point gateways face attacks deploying an ESXi wiper

The campaign began Oct. 30 with attacks on Israeli organizations, then expanded to companies across multiple countries and sectors on Nov. 2, Check Point says. Researchers found political messages in some cases, raising questions about a possible false-flag operation.

A wiper can destroy systems, and this campaign has spread across countries and industries.

[BleepingComputer]

🏠 Apple HomeView puts messages and calendars on display

HomeView goes beyond entertainment, recognizing up to six people and tailoring content across Messages, FaceTime, Calendar, Reminders, and Find My. Available November 13 in select regions, it features a 7-inch Liquid Retina display and Siri AI with onscreen awareness.

For Apple households, it brings messages, calendars and Find My together on a shared home display.

[Apple Newsroom]

🔐 Uber says vendor breach exposed drivers' license data

Hackers used credentials stolen from Fireworks AI to access Uber data held by the vendor, prompting notices to drivers whose personal and license information was taken. Uber learned of the breach on October 9 but hasn’t disclosed how many drivers are affected.

The incident shows how a supplier’s stolen credentials can expose data held for a major service, and Uber hasn’t said how many drivers were affected.

[TechCrunch]

🐛 Vulnerability disclosures pass 100,000 as AI hunting grows

AI-driven vulnerability hunting is fueling a disclosure surge: VulnCheck counted 100,222 CVEs by 4 pm UTC Friday, more than twice 2025’s total. So far, 1.4%—1,363 vulnerabilities—have been added to its Known Exploited Vulnerabilities catalog.

The raw total isn’t a measure of immediate danger: only 1.4% are listed as known exploited vulnerabilities.

[www.theregister.com – Articles]

Also worth knowing

This article was generated by AI and cites original sources.

Scroll to Top