AI Daily brief · 2026-10-21 · WednesdayAI Daily
NEXGATEHUB AI NEWS
US orders federal agencies to remove DeepSeek products and services
Today's most notable developments come from both policy and infrastructure: citing national security, the U.S. Department of Homeland Security has required federal civilian agencies to stop purchasing and remove DeepSeek-related products and services, with the ban also covering third-party products built on its technology. Meanwhile, the PyTorch ecosystem saw dense updates: 2.15's MPS backend brings FP8 to Mac, MoE decoding is 5–6.5x faster, and PyTorch adds native TPU support via TorchTPU.
Model releases/updates
Product releases/updates
MPS in PyTorch 2.15
The Hugging Face PyTorch team introduces the MPS backend updates in PyTorch 2.15: new torch.float8_e4m3fn FP8 support and F.scaled_mm.
Hugging Face ❤️ TorchTPU: TPUs, Natively
Hugging Face announced that, via TorchTPU as a native PyTorch backend, Transformers, Diffusers, and Kernels can now run natively on TPUs without torch_xla or Optimum TPU — models only need device_map="tpu" or pipe.to("tpu").
AWS Machine Learning BlogFirst-partyStructured outputs for Anthropic Claude models on Amazon Bedrock
Amazon Bedrock now supports structured outputs for all Anthropic Claude models, available on the bedrock-runtime and bedrock-mantle endpoints, using constrained decoding during generation to enforce a JSON schema and eliminating the need for custom parsing and retry logic.
Industry news
TechCrunch · AIUS orders federal agencies to remove China’s DeepSeek AI from their systems
The US Department of Homeland Security has ordered federal civilian agencies, citing national security concerns, to stop purchasing DeepSeek products and services and to remove them from government IT systems. The ban also covers third-party products built on DeepSeek technology, and agencies may request exceptions.
Research papers
FairStress: A deeper evaluation of fairness in LLM decision-making
Ai2 released the FairStress benchmark, expanding 300 two-candidate decision scenarios into over 13 million test cases to measure model fairness in hiring, healthcare, lending, and other domains, along with a 50K-item FairStress-Core subset.
ALTK-Evolve: How learned memory tops the τ²-bench leaderboard on customer service tasks without a frontier model
IBM Research released the ALTK-Evolve method, which simply appends behavioral guidelines distilled from the agent's own trajectories to the system prompt, enabling a GPT-4.1-based ReAct customer service agent to reach Pass^4 67.5% on τ²-bench retail.
Tips & opinions
GitHub Blog · AI & MLFirst-partyHow 3 Java Champions are using GitHub Copilot for increased productivity
GitHub Blog interviews three Java Champions with over a decade of experience, who share how they use GitHub Copilot in production and open source projects. Vlad Mihalcea used Copilot to shrink a Hypersistence Optimizer web app from weeks of work to two days, and created a skill based on his 65 articles for documentation.