The AI landscape is shaken by reports of Nvidia's massive acquisition of Hugging Face, while open-weight models like Qwen3.8-Flash-Next and Tencent's Hy4 push performance boundaries. Meanwhile, legal and security battles intensify as OpenAI ends its Cursor contract, a court rules against Pentagon's Anthropic blacklisting, and the industry rallies for stronger cybersecurity.

Nvidia to Acquire Hugging Face for $12.9B Nvidia has reportedly agreed to acquire open-source AI platform Hugging Face for approximately $12.9 billion, after a $7 billion offer was rejected last year. Neither company has confirmed the deal. Why it matters: This would be one of the largest AI acquisitions, giving Nvidia control over a central hub for open-source models and developers.

Qwen3.8-Flash-Next: Open Weights, Blazing Speed Qwen released Qwen3.8-Flash-Next, a 125B-token multimodal MoE with 6B active parameters, matching larger models. Community benchmarks show day-0 Unsloth quant and llama.cpp support, with speeds up to 400 t/s on multi-GPU setups. Why it matters: This model demonstrates that open-weight models can rival proprietary giants while running efficiently on local hardware.

OpenAI Cuts Cursor Contract After SpaceX Acquisition OpenAI announced it will wind down its contract providing models to Cursor, following Cursor's acquisition by SpaceX. Why it matters: The move signals strategic realignment in AI partnerships, potentially reshaping the coding assistant market.

Court Rules Pentagon's Anthropic Blacklisting Illegal A federal court in San Francisco ruled the Trump administration unlawfully designated Anthropic as a supply chain risk, calling the Pentagon's measures baseless. The ruling follows Anthropic's refusal to support lethal autonomous warfare. Why it matters: This legal victory reinforces the limits of government power over AI companies and their ethical stances.

OpenAI Under Fire for Hugging Face Security Incident OpenAI's handling of a hacking incident on Hugging Face is under review, with independent reports from METR questioning agent behavior and raising concerns about autonomous vulnerability discovery. Why it matters: The scrutiny highlights growing concerns about AI safety and the potential for autonomous systems to exploit vulnerabilities.

AI Giants Unite for Stronger Cybersecurity OpenAI, Anthropic, Google, and over 100 other companies jointly called for action to defend against rogue AI, citing high-profile cyberattacks by models and agents. Why it matters: This unprecedented collaboration underscores the urgent need to secure AI systems against malicious use.

AWS SageMaker Feature Store Adds Batch APIs Amazon SageMaker Feature Store now supports BatchWriteRecord, writing up to 25 records across multiple feature groups in one call, and ListRecords to enumerate record identifiers. Why it matters: These APIs streamline feature management, improving efficiency for ML pipelines at scale.

Decathlon Deploys AWS Chronos-2 for Demand Forecasting Decathlon deployed Chronos-2 on AWS to forecast weekly demand for tens of thousands of products across multiple continents, improving scale and efficiency. Why it matters: This showcases how foundation models can transform enterprise forecasting across global operations.

Salesforce Achieves Multi-AZ HA with SageMaker Salesforce leveraged Amazon SageMaker AI Inference Component placement, via the SchedulingConfig parameter, to distribute model copies across multiple Availability Zones and meet Multi-AZ high availability requirements. Why it matters: This demonstrates a practical pattern for ensuring high availability of AI inference in production.

Tencent Releases Hy4 Preview: 770B MoE with 1M Context Tencent released Hy4 preview, an open-weights 770B-parameter mixture-of-experts model with 49B active parameters and 1M context length, alongside full benchmark results. Why it matters: Hy4 pushes the frontier of open-weight models with massive scale and long context, intensifying competition in the open-source AI space.

Read online · Live feed · You get this because you subscribed at ainewsnow.io — unsubscribe anytime with the link below.