– Google’s TPUv7, used to train frontier models like Gemini 3 and Claude 4.5 Opus, is now sold directly to external customers, including a landmark deal with Anthropic for up to 1 million chips.
– NVIDIA’s GB200 NVL72 rack-scale system enables mixture-of-experts (MoE) AI models to run 10x faster by leveraging 72 Blackwell GPUs, high-bandwidth memory, and a unified NVLink interconnect.
– Mistral launched Devstral 2, a 123-billion parameter coding model, and Devstral Small 2 at 24B parameters, both available for free via API and Hugging Face for a limited time.
– Cursor, the AI coding assistant, reached $1 billion in annualized revenue in November 2025 and raised $2.3 billion at a $29.3 billion valuation.

– Mistral AI launched Devstral 2, a new coding-focused AI model, and introduced Mistral Vibe, a command-line interface for code automation, alongside a smaller version, Devstral Small.
– Nvidia received U.S. approval to sell its H200 AI chips to approved commercial customers in China, with a 25% tariff imposed on these sales.
– OpenAI announced a collaboration with Deutsche Telekom to bring advanced AI capabilities to millions across Europe, including multilingual AI experiences and the rollout of ChatGPT Enterprise to Deutsche Telekom employees.
– Google launched a new AI Quest to promote AI literacy in schools, backed by over $5 million in Google.org funding and expanded partnerships for computer science education.

– IBM will acquire Confluent in an all-cash deal valued at $11 billion, paying $31.00 per share, to enhance real-time data streaming capabilities for AI models.
– Google plans to launch its first AI-powered smart glasses with Gemini in 2026, collaborating with Samsung Electronics, Warby Parker, and Gentle Monster.
– Anthropic has launched a beta integration of Claude Code with Slack, enabling agentic coding tasks like bug fixes and feature implementations directly within Slack threads.
– OpenAI and Instacart have partnered to integrate the Instacart app into ChatGPT via the Agentic Commerce Protocol, allowing users to browse, add, and securely checkout groceries within the chat interface.

– GPTZero identified 50+ verified hallucinations in 300 ICLR 2026 conference submissions, including fabricated authors and false citations, despite average peer review ratings as high as 8/10.
– Berkeley AI Research deployed 100 reinforcement learning-controlled autonomous vehicles on a highway to smooth stop-and-go traffic waves, reducing fuel consumption and improving flow using local sensor data and fast simulations.
– PLAID, a multimodal generative model from Berkeley AI Research, generates valid protein sequences and 3D structures by leveraging the latent space of ESMFold, using only sequence data for training.
– LinkBERT, developed by Stanford AI Lab, improves language model pretraining by incorporating document-level links, enhancing contextual understanding in models like BERT and GPT series.

– Netflix won a $72 billion cash and stock deal to acquire Warner Bros. Discovery, beating competitors including Paramount Skydance Corp. and Comcast.
– Google launched Gemini 3 in November 2025, now available in Google Search and the Gemini app, with the introduction of Nano Banana Pro for high-fidelity image generation and editing.
– Google announced a $40 billion investment in AI and cloud infrastructure in Texas, set to expand its AI capabilities and data center capacity.
– Jeremy Carrasco, a former producer and director, gained over 300,000 followers on both TikTok and Instagram by sharing AI literacy content focused on identifying AI-generated videos.

# AI Daily Roundup: December 5, 2025

—

## TL;DR

– [The New York Times sued Perplexity for copyright infringement](https://techcrunch.com/2025/12/05/the-new-york-times-is-suing-perplexity-for-copyright-infringement/) over alleged “verbatim copies” of its content.
– [Meta announced commercial dat…

– Today brought several significant AI developments across the industry
– Multiple companies announced updates to their AI products
– The community continues to push boundaries in research and applications