Daily summary · 25 October 2026

The most relevant AI news of the last 24 hours. Each one links to its story and to the original source.

1. Image & Video Gen

   Canva launches its first in-house image model, Canva Craft

   Design platform Canva has released its first in-house image generation model, Canva Craft. It is available to all users, promising better, more customizable performance at lower costs.

   Source: The Decoder, The Verge, TechCrunch
2. Agents & Dev

   Claude Code v2.1.310 released with Monitor tool addition and bug fixes

   Claude Code v2.1.310 has been released. It adds the Monitor tool and plugin monitors for Bedrock, Vertex, and Foundry, alongside several fixes for request failures, file deletion, and memory usage.

   Source: Claude Code
3. Local & Open Source

   Ollama v0.40.11 released with speech generation support on macOS

   Ollama v0.40.11 has been released, featuring speech generation on macOS and the introduction of d3 decision models from vLLM Semantic Router.

   Source: Ollama
4. Policy & Society

   Google suggests Gmail and other data may be used for ads when using personalized AI features

   Google has updated its privacy policy. If Personal Intelligence is enabled, data from Gmail and Google Photos may be used for ad targeting.

   Source: Ars Technica, The Verge
5. Hugging Face

   deepseek-ai/DeepSeek-V4.1-Flash-1017 trends on Hugging Face

   The model deepseek-ai/DeepSeek-V4.1-Flash-1017, designed for image-text-to-text tasks, is trending on Hugging Face.

   Source: Hugging Face Trending
6. Workflows & Nodes

   Moving "one illustration" with RX 9060 XT × Windows — Actual measurement records of pose correction, 10-second video, and AI inspection

   This is an actual measurement record of changing the pose of one anime-style illustration, turning it into an approximately 10-second AI video, and inspecting the resulting image and video on a Windows PC equipped with AMD's graphics board "RX 9060 XT" (VRAM 16GB).

   Source: Zenn (ComfyUI), Zenn (Stable Diffusion)
7. AI Companies

   Three fired OpenAI safety researchers deny allegations of leaking confidential information

   Fired researchers Tomek Korbak, Mikita Balesni, and Jasmine Wang deny sharing confidential information and claim OpenAI never informed them in writing what was allegedly leaked.

   Source: The Decoder
8. AI Companies

   Spotify takes measures against "low-quality AI music," excluding 50 million songs from recommendations

   Spotify disclosed the progress of its efforts to protect artists and listeners, such as measures against AI-utilized spam and slop (low-quality generative AI content). In 2026 alone, it has excluded more than 50 million spam/slop songs from recommendations and introduced features such as identification of AI-generated artists and disclosure of AI usage.

   Source: Impress Watch

Get this summary by email every morning →

← Previous summaries