Daily Summary of AI · Final editionSunday 15 November 2026 · UTC
AI News: 15 November 2026
Anthropic IPO nears as Congress seeks tighter AI chip export rules
Anthropic is expected to debut on the Nasdaq on Thursday at an IPO price of $70 to $78, with major tech firms investing $26 billion. In Congress, Democrats and Republicans are backing measures to restrict AI chip exports to China after a suspected Chinese spy scandal involving former Nvidia CEO Jensen Huang.
Anthropic's Claude Fable 5.2 can create paintings by writing code that places simulated brushstrokes, showing a different route to image generation.
Congress seeks tighter AI chip export rules after Nvidia spy scandal
Following a suspected Chinese spy scandal involving former Nvidia CEO Jensen Huang, Democrats and Republicans in Congress are backing measures to restrict AI chip exports to China. Lawmakers have demanded that Huang and other Nvidia leaders testify, and the Remote Access Security Act, introduced in the House in April 2025 and stalled in the Senate, is receiving renewed interest.
Anthropic's Claude Fable 5.2 paints with simulated brushstrokes
Anthropic's Claude Fable 5.2 can create paintings by writing code that places hundreds of thousands of simulated brushstrokes on a virtual canvas. Unlike diffusion image models, it simulates bristle texture, wet paint and canvas texture, and can paint over mistakes. Developer Jake Eaton showed a shopping cart image made with roughly 700,000 strokes, showing a different route to image generation.
Anthropic's IPO nears as AI trade seeks a bright spot
Anthropic is expected to debut on the Nasdaq on Thursday at an IPO price of $70 to $78, with major tech firms investing $26 billion. The company has outstanding spending commitments of over $900 billion, including at least $110 billion with Amazon over a decade and $125 billion to $150 billion for Google AI chips. It recorded a net loss of $104 billion through the first nine months of 2026, and the IPO could raise over $90 billion.
AI DJs in music apps draw attention for intrusive, error-prone commentary
AI DJ features are drawing attention as listeners encounter synthetic commentary between songs. YouTube Music's Beyond the Beat, launched in 2025, inserts about 30 seconds of context before tracks, while Spotify expanded its AI DJ personalities and controls in November. The reporter found the commentary intrusive and error-prone, including a mispronunciation of Charli XCX and a mismatch between the description and the track played.
Megaspeed brokered nearly $12 billion in Nvidia chip orders for China
WSJ-reviewed documents show Megaspeed arranged nearly $12 billion in Nvidia chip orders for Chinese companies from March 2024 to June 2025. The company, owned by Alice Huang, shipped several hundred thousand export-controlled chips to buyers with ties to Chinese tech giants. Nvidia reported $12.4 billion in China revenue over six months, and Culper Research estimates over 20 percent of its last fiscal year revenue came from Chinese customers.
AI Tamagotchi OATs turn everyday objects into needy virtual pets
OATs, or Object Agent Twins, are micro-agent AI avatars on E Ink displays that pair with household objects. Created by Carnegie Mellon PhD student Daehwa Kim, the project was presented at the ACM Symposium on User Interface Software and Technology and earned two Honorable Mention awards. The devices use sensors for moisture, temperature, humidity, light, occupancy and motion, with sprites generated by Google Gemini's Nano Banana.
Australia's central bank finds AI hitting young workers first
Australia's central bank analysis found job postings have fallen since ChatGPT launched in late 2022, with the steepest drop in highly AI-exposed occupations such as programming, clerical work and customer service. In that group, young workers aged 15 to 24 and recent graduates have weaker employment rates and higher unemployment than older workers. The study does not prove causation, but similar findings point to pressure on junior roles.
Bland launches Tofu-1, a full-duplex speech-to-speech model
Bland released Tofu-1, a full-duplex speech-to-speech model in research preview that hears, reasons and replies in audio during phone calls. It processes audio directly, allowing callers to interrupt or correct it mid-sentence while it works on a request. The company is offering limited enterprise access and targets banking, healthcare, insurance and government.
Meituan releases LongCat-2.5, a 1.6-trillion-parameter agent model
Meituan released the weights for LongCat-2.5, a 1.6 trillion parameter mixture-of-experts model built for autonomous agents, with about 48 billion parameters active per token. The company said it was trained entirely on Chinese hardware and performs near top Western models on coding benchmarks. The model accepts text, images and video, handles up to one million tokens at once, and can generate responses up to 128,000 tokens.
South Korea's top tax exam candidate admits AI cheating before college exam
Han Won-seok, who scored 397.5 out of 400 in the first round of South Korea's certified tax accountancy exam, admitted cheating with a microcamera, smartphone and generative AI. He said he photographed questions, sent them to an AI tool and checked answers on his phone. The confession, days before the national college entrance exam, prompted emergency meetings and a five-year retake ban for those found to have used AI.