2026-10-20 · TuesdayPublished daily at 08:00
AI Daily
The U.S. Department of Homeland Security ordered all federal civilian agencies to remove DeepSeek and products developed based on it

The U.S. Department of Homeland Security issued Order 2026-01 under the 2018 supply chain law, requiring all federal civilian agencies to immediately stop purchasing products from DeepSeek and its affiliates and remove related products from their IT systems, including third-party products and services developed using DeepSeek implementations.
- 1
- 2Ant's inclusionAI open-sources the Ming-3.0-tts speech synthesis modelAnt Group inclusionAI: New GitHub Repository
- 3
01Model Releases/Updates
MiniMax released M3.1 Flash, nearly tripling inference speed and reducing usage costs by 70%
MiniMax releases MiniMax M3.1 Flash. The anonymous model Space Bunny was an early version and had processed more than 200 trillion tokens.
Original textAnt's inclusionAI open-sources the Ming-3.0-tts speech synthesis model
Ant Group's inclusionAI releases Ming-3.0-tts on GitHub, including two speech-generation models: a 0.5B dense model and a Ling V3-based tiny model with 7.9B total parameters and 1.3B active parameters. They are open-sourced under the MIT license, with weights available for download from Hugging Face and ModelScope.
Original textfal and bfl_ai Release FLUX 3 Max, a Post-Trained Video Model
fal releases its latest post-trained video model FLUX 3 Max in partnership with @bfl_ai. It tied for first in third-party image-to-video quality evaluations and takes about 12 seconds to generate a 10-second 720/768p video. The model will be offered at a 75% discount for the next three weeks and is described as the least expensive frontier video model in its quality tier.
Original text02Product Releases/Updates
Inception's diffusion model Mercury 3 launches on OpenRouter
Inception's diffusion model Mercury 3 launches on OpenRouter, generating at over 1,000 tok/s and supporting a 1M context, four times that of Mercury 2.5. Inception reported that it surpasses GPT-6 Luna and Haiku 5.5 on IFBench, GPQA Diamond, and SWE-Bench Atlas; during the launch period, it is offered at half price, at $0.10/M for input and $0.35/M for output.
Original textAWS releases TorchNeuron: a native PyTorch backend for Trainium
AWS released TorchNeuron, an open-source native PyTorch backend for AWS Trainium. It will enter public beta under the Apache 2.0 license on October 20 at PyTorch Conference North America. The code is available at github.com/aws-neuron/torch-neuronx.
Original textClaude launches two new tools, Dashboards and Motion, and graduates Docs, Slides, and Design from beta
Claude announced that Dashboards and Motion are entering beta. Dashboards can connect to data platforms or CRMs to generate dashboards that update in real time as data changes, and is available on all paid plans. Motion turns reports or charts into short animated explainer videos generated through code, without video models. Users can edit the text, numbers, and pacing before exporting an MP4. It is available on Team and Enterprise plans.
Original text03Industry News
The U.S. Department of Homeland Security ordered all federal civilian agencies to remove DeepSeek and products developed based on it
The U.S. Department of Homeland Security issued Order 2026-01 under the 2018 supply chain law, requiring all federal civilian agencies to immediately stop purchasing products from DeepSeek and its affiliates and remove related products from their IT systems, including third-party products and services developed using DeepSeek implementations.
Original textHackers use AI agents to breach over 100 websites, including John Deere and SharkNinja
AI security company Norforge discovered that a hacker used agents orchestrated with the open-source AI platform Cairn to breach more than 100 websites. According to Reuters, victims included John Deere, SharkNinja, and Cummins.
Original text04Paper research
OpenAI releases MultiLangBench and a COBOL migration benchmark, with the GPT-6 series showing major improvements in low-resource programming languages
OpenAI announced significant improvements in its latest models' ability to build and migrate software in low-resource programming languages such as COBOL, Fortran, and Visual Basic, and open-sourced an evaluation suite.
Original text05Tips and perspectives
Meta Explains Its Gigawatt-Scale Prometheus Supercluster in Detail: Phase One Completed and Used to Train the Muse Model
Meta introduces Prometheus, its gigawatt-scale supercluster in New Albany, Ohio. Phase one is complete and being used to train the Muse series of foundation models, with a target expansion to approximately 1 million GPUs.
Original textMeta explains the Thin Orchestrator pattern in detail: Make routing do less to improve AI agent reliability.
Meta's engineering team describes the thin orchestrator pattern used by its data-pipeline agents: an orchestrator of approximately 1,300 lines is responsible only for routing and reliability checks, dispatching more than 20 workflows to 14 specialized subagents with isolated contexts, while performing no domain reasoning itself.
Original textn8n Tutorial: Run a Complete AI Workflow Offline on NVIDIA DGX Spark
The n8n blog published a tutorial explaining how to deploy seven local models on a single NVIDIA DGX Spark (128GB unified memory), use LiteLLM’s single OpenAI-compatible endpoint to drive n8n workflows, and avoid calling cloud APIs throughout.
Original text06News Briefs
- DeepSeek V4.1 Flash 1017 released, scoring 45 on the Artificial Analysis Intelligence Index and tying for second among open-source modelsArtificial Analysis
- FBI says former OpenAI engineer is suspected of stealing secrets; Saudi Gov Docs folder found on phoneArs Technica:AI
- Disney plans to acquire the Character.AI team and technology through bankruptcy proceedings, bypassing its lawsuitArs Technica:AI
- Google ADK Tutorial: Building Voice Agents That Never Run Out of Conversation with Streaming Tools and Response SchedulingGoogle Developers Blog
- How Numan built the evaluation-first AI health coach Nu with LangGraph and LangSmithLangChain:Blog
- Meta Muse Wonder Full (Max) ranks sixth on Code Arena: WebDev with 1,726 points.Arena
- How Microsoft uses Project Perception to protect against agentic insiders within enterprisesMicrosoft:Official Blog
- Grok Bot adds support for the iPhone Action buttonGrok Bot