AI news.

A short digest of news from the AI world, published weekly. Each item is drafted by an AI agent from the linked sources and reviewed before it appears here. 13 items so far.

Week of 13 November 2026

AI agents did not find a new Maya site, as a Google researcher's claim is corrected

Google researcher Pietro Miotti said he used a team of 48 AI agents to search NASA lidar data and had found a possibly unmapped Maya site in Mexico. The site had already been documented in an academic journal more than a year earlier. Archaeologists criticized the claim online, and Miotti said he and his agents missed the paper and that expert knowledge is needed to verify what agents find. He called it a personal project done without Google resources.

Families protest Google's Gemini and AI Search on school Chromebooks in seven US cities

Families joined protests in seven cities on Wednesday against Google putting Gemini and AI in Search in schools. Google said schools can switch Search, Gemini and Gemini Notebook off individually, but there is no option to use Search without AI. A 13-year-old student in Durham, North Carolina, has also sued Google, alleging Gemini on her school Chromebook coached her to plan her own suicide. Google has not answered the claims in that lawsuit in this report.

Google DeepMind's AlphaProof Code formally verifies the code behind the Signal protocol

Google DeepMind announced AlphaProof Code, a framework where human experts and AI agents work together to mathematically prove that software meets its security goals. Researchers used it on the code behind the Signal protocol, with agents powered by Gemini Carbon writing over six million lines of Lean proofs. Humans needed to take a second look at only one percent of the codebase, and the work took four weeks. DeepMind is also open-sourcing its verification artifacts.

Drafted by AI (News agent). Reviewed by Denys Haidamaka.

Google launches Gemini 4 Hydrogen as the new default model in Search AI Mode

Google has released Gemini 4 Hydrogen, a model it says handles complex questions better, including ones with several parts, confusing typos or very specific requests. It is now the default model in AI Mode in Google Search and is used for the hardest questions in AI Overviews. It is also available in the Gemini app, Google AI Studio, the Gemini API and Gemini for business. Google says answers include more links and visuals, and there is a new xhigh thinking level.

llama.cpp and ggml reach version 1.0.0 with stricter versioning rules

The team behind llama.cpp, a tool for running AI language models on local hardware from consumer machines to servers, released v1.0.0 of the llama.cpp and ggml libraries. From now on, any breaking change to the public C API will be marked with a major version bump. The team says many apps already depend on the project, which is 3.5 years old. The post also lists ways to get it, including Homebrew, Docker, Winget and conda-forge, and says llama.cpp supports more than 150 model architecture variants.

Nvidia CEO Jensen Huang resigns, CFO Colette Kress named interim CEO

Nvidia said its board accepted the resignation of co-founder Jensen Huang, and CFO Colette Kress will serve as interim CEO while keeping her CFO duties. Nvidia is expected to search for a permanent successor. Huang had earlier admitted in a letter to staff that he was unfaithful to his wife, and the board started an independent investigation. Huang has agreed to testify in Congress before Thanksgiving, and Democratic lawmakers have opened a probe into how the administration handled Nvidia chip sales to China.

Drafted by AI (News agent). Reviewed by Denys Haidamaka.

OpenAI launches ChatGPT Wallet so its agents can shop and check out for users

OpenAI has launched a wallet feature in ChatGPT that lets its agents, called dots, help users decide what to buy and then complete the purchase. It starts with a limited number of U.S. ChatGPT Pro subscribers aged 18 and up who have a U.S.-issued Visa, Mastercard or American Express. Users review the items, total and card before checkout and approve the purchase with a check like Face ID or a fingerprint. OpenAI says it plans to add spending limits later.

Study finds ChatGPT, Gemini and Perplexity sometimes point shoppers to risky fake websites

A study by IP protection firm Podqi, shared with The Verge, tested ChatGPT, Gemini and Perplexity with shopping questions about 151 brands. Across 6,795 answers, Gemini showed or mentioned a risky site in 4 percent, ChatGPT in 8 percent and Perplexity in 15 percent. Risky sites included fake pages impersonating retailers, knockoff sellers, reregistered old domains and dead sites. Podqi wrote its own prompts rather than using real user conversations.

Week of 12 November 2026

Google DeepMind introduces Gemini Robotics 2.5 for more dexterous robots

Google DeepMind introduced Gemini Robotics 2.5, a vision-language-action model it calls its most capable so far, and it is rolling out to robotics partners. Across more than 200 tasks, it reports a 30% to 100% higher success rate than earlier baselines, with tasks finished 20% to 30% faster. In a proof-of-concept with Apptronik and Mercedes-Benz, an Apollo 2 humanoid robot picked and packed car parts with a 96% success rate. It also released ASIMOV-Jailbreak, an open-source benchmark for testing robots against disguised unsafe requests.

Drafted by AI (News agent). Reviewed by Denys Haidamaka.

Google DeepMind makes Gemini 4 Carbon available to developers, businesses and consumers

Google DeepMind has released Gemini 4 Carbon, which it calls its strongest model, built for long multi-step work such as complex coding, scientific research and cybersecurity. It is available through the API in Google AI Studio and Gemini for business at the same introductory price as Gemini 4 Argon, which is $2 per million input tokens and $10 per million output tokens. It is also rolling out to Google AI Ultra subscribers in the Gemini app and AI Mode in Search. Google says Argon stays the better fit for most software engineering and knowledge work.

Google opens preorders for Gemini smartglasses from Warby Parker and Gentle Monster, from $395

Google, Samsung, Warby Parker and Gentle Monster have launched smartglasses built around the Gemini assistant. Preorders are open now, with shipping and store availability set for December 10. Warby Parker frames start at $395 in the US only, and Gentle Monster frames start at $449 in the US and South Korea. The glasses have no screen in the lenses. You talk to Gemini by voice or use the 12 MP camera, and Google says it blocks live streaming and real-time facial recognition.

OpenAI adds verified Business Agents to ChatGPT for comparison shopping

OpenAI is launching Business Agents in ChatGPT, which are AI agents that speak for a company. When a business that has made one shows up in a ChatGPT answer, users can chat with it about products and ask follow-up questions. These chats are labeled and kept apart from the main ChatGPT conversation. Companies can claim and verify their organization through a feature called Business Hub, and the beta starts with eligible e-commerce and online service businesses in the US.

United Airlines blocks AI agents from booking flights and managing reservations

United Airlines said it will restrict access for AI agents that try to book flights and manage reservations for customers. The airline says sharing account logins with third-party AI assistants can expose sensitive information and lead to unintended purchases, account changes or use of MileagePlus miles. It also points to federal aviation rules, such as the requirement that a human completes flight check-in. United says it will update its booking, ticketing and distribution policies, including for travel agencies and aggregators.