← Tuesday 27 Octobernote 4 of 13
Open-weight AI modelsTue 27 Oct 2026 · 4 of 13

Growth of Chinese Open-Weight AI Model Use Raises Security Scrutiny

Unsupported claims · 5/6 supported1 cited · 1 min read

cohere.com reports that companies and government contractors adopt open-weight AI models believing that running them on their own servers keeps data secure . cohere.com reports that security officials in Ottawa, Washington DC, London, Berlin and Tokyo see two options: closed-API services or self-hosted open-weight models . cohere.com reports that thousands of engineering teams have chosen to self-host open-weight models, increasingly using Chinese models . cohere.com reports that Chinese models accounted for 41 % of downloads on Hugging Face in the year before March 2026, and handled more than 30 % of AI work on OpenRouter each week from February through early July 2026 . cohere.com reports that U.S. lawmakers opened an investigation into corporate use of Chinese AI models in spring 2026 .

Why it matters. cohere.com reports that the investigation means U.S. authorities are scrutinizing corporate reliance on Chinese AI models, which could affect companies that self-host them .

Earlier on Open-weight AI models

Fri 23 Oct 2026Hugging Face launches OlmoEarth v1.3 with global embeddings and web explorerHugging Face released OlmoEarth v1.3, a state-of-the-art satellite embedding model. The model provides precomputed annual embeddings for the entire planet at 10‑meter resolution in both 64‑ and 128‑dimensional formats. A new browser‑based OlmoEarth Embeddings Explorer lets users explore and customize maps without sign‑in or advanced technical skills. OlmoEarth v1.3 reads a year of Sentinel‑1 radar and Sentinel‑2/Landsat multispectral time series to produce a 128‑dimensional vector for each 10‑meter pixel, with the first 64 dimensions usable independently. This embedding approach replaces the previous fine‑tuning workflow that required GPU access and imagery download, enabling cheap, reusable features for many tasks.Thu 22 Oct 2026NVIDIA Kumo Relational Model Outperforms Supervised Methods on Relational BenchmarksNVIDIA Kumo Relational is a pre‑trained relational foundation model that makes predictions directly on enterprise databases. For a predictive query it gathers up to 10,000 historical examples as context and predicts outcomes such as churn, demand, or fraud in a single forward pass. The model requires no task‑specific training, no feature engineering, and no flattening of tables. It is the first few‑shot foundation model to outperform supervised machine learning on common relational benchmarks, beating supervised relational deep‑learning models and expert‑engineered gradient‑boosted tree pipelines across 41 tasks on 15 databases, and surpassing AutoGluon ensembles by 8% in mean reciprocal rank on the SAP SALT benchmark. Kumo Relational scales to databases with over 500 billion rows and is available through NVIDIA’s open‑source GPU‑native structured‑data‑models library or as an NVIDIA NIM.Wed 21 Oct 2026Hugging Face launches TorchTPU for native PyTorch support on Google Cloud TPUsHugging Face introduced TorchTPU, a native PyTorch backend for TPUs. With TorchTPU, Transformers, Diffusers and Kernels run natively on TPUs within the Hugging Face ecosystem. TPUs are Google’s AI accelerators built for matrix math and are used by Google to train and serve models such as Gemini. Google Cloud offers TPUs in configurations ranging from a single chip to pods of thousands of chips. Previously, PyTorch on TPUs required torch_xla with lazy tensors and Optimum TPU for Hugging Face models, but those are no longer needed with TorchTPU.
Next · Health researchHeme shown to inhibit clot‑dissolving enzyme, study finds →