White House OSTP Director Michael Kratsios says Moonshot AI "acquired GB300-equipped servers and has accessed GB300s in Thailand, likely to train its AI models"
THE SO WHAT
Allegations that Moonshot trained on GB300s via Thailand underscore how export controls are now a routing problem, not just a licensing one. Multinationals training or serving models across borders need a live map of where chips, models, and data actually sit — regulators are watching the full supply chain, not just your HQ.
READ THE SOURCE
MORE FROM THE WIRE
Applied AIMonday.com lays off hundreds to focus on AI
Cutting ~20% of headcount to fund an "AI Work Platform" pivot is a clear signal: horizontal SaaS is repricing around AI-native value, not incremental features. If you compete in work management or collaboration, assume your buyers will ask what you’re cutting to fund AI—not just what you’re shipping.
'Gemini who?': Rivals dunk on Google's delayed frontier AI
Frontier model timelines are now part of competitive positioning — a 2026 slip moves you from ‘pace-setter’ to ‘catch-up’ in the narrative, whether or not the tech gap is real. If you’re building on a single vendor’s roadmap, assume slippage and architect for model interchangeability now.
Applied AIMedtronic to launch AI compute platform for the operating room
Medtronic putting dedicated AI compute into the OR—spanning pre-op planning, intra-op tele-mentoring, and post-op analytics—turns surgery into a continuous data pipeline. Hospital leaders should treat the OR like an edge compute site and get ahead of data governance, latency, and vendor lock-in questions now.
Applied AIOpenAI Models Breach Hugging Face, Sparking Cyber Alarms
Advanced models are now both the attack surface and the asset at risk — the Hugging Face incident underscores that model supply chains are security-critical infrastructure. Treat your model hosting, weights, and fine-tunes with the same controls you’d apply to source code and production data, not as a sidecar dev tool.