12:28 AM
Today in brief.
Canva builds its own image model
Canva launched Canva Craft, its first in-house image model, which plans a design before generating it, at up to 2K and 30+ aspect ratios.
Claude proves security properties in 100+ libraries
Anthropic and Theorem Labs used Claude to formally verify security properties in over a hundred real open-source libraries in a few weeks, while PHP2Shell attackers turned out to have a four-day head start.
Decision models go local and go batch
Perplexity's Decision Model now runs fully on a Mac, OpenAI opened its Jev competitor widely, and Perplexity's Agent API now takes 50k requests per batch file.
Polls sour on AI while companies go all in
An ABC News/Ipsos poll finds most Americans say AI is bad for society and oppose more data centers, the same day Uber put all 30,000 staff on ChatGPT Work and Barclays said it will cut 5,000 jobs.
OpenAI lets METR and Redwood inside its monitoring
OpenAI gave METR and Redwood Research three weeks of deep infrastructure access to evaluate its internal monitoring system, as OpenAI, Anthropic and Nvidia posted record lobbying quarters.
Try today: forest maps, Pokemon agents, AI island games
DeepMind released a 10m map that separates natural, planted and farmed forests, LLM agents now battle in Pokemon Showdown, and MIT marooned 13 AIs on a virtual island where they invented games.
Microsoft's AI beats doctors on hard cases, with a catch
A Science study says Microsoft's diagnostic system beat generalist doctors on hard cases, but most of the gain comes from the models underneath, and a separate CT study flagged coronary inflammation in 42.9% of patients.
Agents get logins, a sandbox and a shared skills standard
1Password lets agents log in for you without seeing your password, Anthropic donated Agent Skills to the Linux Foundation, and Nvidia open-sourced OpenShell, a private runtime for agent fleets.
OpenAI eyes $1.25T, DeepSeek raises $15B
OpenAI is working on a $30B raise at a $1.25T valuation ahead of an IPO, and DeepSeek reportedly raised $15B days after the U.S. ordered agencies to drop its tech, as DeepSeek and MiniMax cleared China's export rules for overseas open-source releases.
No stories in this category for this edition.