Why it mattersIt is Europe's most ambitious bid to stay in the frontier race with open weights, pitched at cybersecurity work that closed US models often refuse.
Mistral's Large 4 is the company's biggest model yet, with one trillion parameters trained on its own European infrastructure. In the independent Intelligence Index, the model makes a big leap forward but still falls well short of Claude, GPT-6, and Chinese competitors. — The Decoder
With the release of its new trillion-parameter model, Mistral is hoping to demonstrate it’s “still in the race” to build frontier-level artificial intelligence. — Wired
Open models are vying to become the everyman defense solution for a growing swarm of agent breaches. — ZDNET
Common Sense Media rated OpenAI's ChatGPT for Teens an “unacceptable risk,” saying it doesn't alert parents when it should, falls short in crisis situations and still does kids' homework.
Anthropic expanded its Claude for Startups program to offer a free year of Claude Team with up to five premium seats, plus $1,000 in API credits. Companies founded in the last five years or funded in the last two are eligible.
ChatGPT now accepts audio uploads, turning recordings of meetings, interviews or lectures into transcripts, summaries, structured notes and follow-up drafts. The feature is available on paid subscriptions and workspaces.
Anthropic folded Project Glasswing and its Cyber Verification Program into one expanded program with three access tiers. Vetted security teams get reduced cyber safeguards on its most capable models, including Claude Mythos 5.1.
Rep. Lori Trahan is rolling out the draft CLAIM Act, which would make it easier for third parties to sue AI developers when their agents cause harm. The bill responds to this year's rogue AI hack of Hugging Face.
Finland's licensing agency ordered Google to halt work on two planned data centers in Muhos and Kajaani until environmental impact assessments are done, a month after Google pledged $15 billion for AI infrastructure there. Google said it had “fallen short of our own high standards.”
GitHub says it is rebuilding its Git infrastructure for agentic development, where developers and agents work concurrently in repositories that receive millions of commits a day, and lays out the design principles behind the work.
Version 1.2.44 of the Grok Android app makes Customize Grok available to all users, fixes table formatting issues and improves sharing and downloading in Imagine.
Cursor's iOS app can now see and reply to local agents running on a user's computer after a one-time pairing with the desktop app. Remote Control is on by default for everyone except Enterprise organizations.
OpenAI put its Decisions API into public beta for checking conditions, choosing among fixed options and scoring text and images against a rubric.
Nvidia-backed Lambda is raising up to $4 billion at a $14.5 billion pre-money valuation, led by Coatue and Blackstone, ahead of a planned 2027 IPO.
The boom in personal AI agents such as Meta's Muse is lifting AMD and Intel, which have outperformed their megacap tech peers over the past month as agent workloads drive CPU demand.
Mistral opened a public preview of Mistral Large 4, nicknamed “le Chonk,” a natively multimodal model with 1 trillion parameters and 49 billion active, trained from scratch on 3,800 NVIDIA Grace Blackwell GPUs in its own European data centers. The preview API is live on Mistral Studio, and the company says the open weights will follow by the end of the month.
Why it mattersIt is Europe's most ambitious bid to stay in the frontier race with open weights, pitched at cybersecurity work that closed US models often refuse.
Anthropic launched a public beta that puts Claude in a sidebar inside Google Docs, Sheets and Slides for all paid Claude plans, alongside beta connectors that create and edit Google files from Claude itself. Edits default to “Ask before edits,” and enterprise controls such as the Compliance API carry over.
Why it mattersIt brings Claude directly into office files many businesses already use, with edits needing user approval by default.
Google DeepMind released EmbeddingGemma 2, an open 740-million-parameter embedding model that turns text, images, video, audio and code into vectors and runs on-device in about 191 MB of RAM. Google says it outperforms some rival models twice its size.
Why it mattersSmall embedding models that run offline make private, on-device search and retrieval practical without sending data to outside servers.
OpenAI published new results on open problems in mathematics from an internal frontier model, posting 372 AI-generated results with Lean formalizations on GitHub. The release stunned mathematicians and drew warnings from 25 Fields Medal winners that mass-produced proofs could crowd out new ideas.
Why it mattersMass-produced, machine-checkable proofs are forcing a debate over what AI-generated results mean for how mathematical research gets done.
Common Sense Media rated OpenAI's ChatGPT for Teens an “unacceptable risk,” saying it doesn't alert parents when it should, falls short in crisis situations and still does kids' homework.
Anthropic expanded its Claude for Startups program to offer a free year of Claude Team with up to five premium seats, plus $1,000 in API credits. Companies founded in the last five years or funded in the last two are eligible.
ChatGPT now accepts audio uploads, turning recordings of meetings, interviews or lectures into transcripts, summaries, structured notes and follow-up drafts. The feature is available on paid subscriptions and workspaces.
Anthropic folded Project Glasswing and its Cyber Verification Program into one expanded program with three access tiers. Vetted security teams get reduced cyber safeguards on its most capable models, including Claude Mythos 5.1.
Rep. Lori Trahan is rolling out the draft CLAIM Act, which would make it easier for third parties to sue AI developers when their agents cause harm. The bill responds to this year's rogue AI hack of Hugging Face.
Finland's licensing agency ordered Google to halt work on two planned data centers in Muhos and Kajaani until environmental impact assessments are done, a month after Google pledged $15 billion for AI infrastructure there. Google said it had “fallen short of our own high standards.”
GitHub says it is rebuilding its Git infrastructure for agentic development, where developers and agents work concurrently in repositories that receive millions of commits a day, and lays out the design principles behind the work.
Version 1.2.44 of the Grok Android app makes Customize Grok available to all users, fixes table formatting issues and improves sharing and downloading in Imagine.
Cursor's iOS app can now see and reply to local agents running on a user's computer after a one-time pairing with the desktop app. Remote Control is on by default for everyone except Enterprise organizations.
OpenAI put its Decisions API into public beta for checking conditions, choosing among fixed options and scoring text and images against a rubric.
Nvidia-backed Lambda is raising up to $4 billion at a $14.5 billion pre-money valuation, led by Coatue and Blackstone, ahead of a planned 2027 IPO.
The boom in personal AI agents such as Meta's Muse is lifting AMD and Intel, which have outperformed their megacap tech peers over the past month as agent workloads drive CPU demand.
No stories for this company in today’s edition.