Australian firm Trellis Data has released Agentaus 1, a 220B-parameter model built around data sovereignty: prompts stay on Australian servers and are never used for training. The company reports 84.4% on MMLU-Pro, ranking 43rd among foundation models, and says it runs on two GPUs where comparable models need four.
Sovereign AI
主权 AI
Paradigm of building AI capacity under national or local control, keeping data and compute within borders.
Curated Intelligence
Intel announced at its Seoul AI Summit the formation of the K-AI Full-Stack Alliance with 23 Korean companies spanning server manufacturing, NPU design, foundation models and cloud services, aimed at building sovereign and open AI infrastructure for the Korean market. Intel also publicly showed its D270 data-center inference accelerator, claiming 26,200 TFLOPS of FP4 compute at 350W and advantages over Nvidia's RTX 6000 Pro in Gemma latency, agent performance and financial simulation.
NVIDIA announced an expanded sovereign AI collaboration with Korea, extending from GPU compute infrastructure to local AI talent and skills development. The announcement is a vendor-led industry partnership without specific model parameters, technical methods, or verifiable deployment metrics.
French AI company Mistral has released the weights for its trillion-parameter open-source model, Mistral Large 4 (ML4), enabling self-deployment by enterprises and governments. Nicknamed 'Le Chonk,' the model claims state-of-the-art performance at a fraction of the compute cost of closed labs, with a specific focus on cybersecurity applications requiring data sovereignty.
France is accelerating the deployment of sovereign AI infrastructure using the NVIDIA stack, spanning from ground data centers to space-based edge computing. Bull announced plans to extend European AI server production to NVIDIA Vera Rubin Ultra NVL72 systems by 2028, while the Altair-1 satellite with onboard autonomous capabilities has successfully reached orbit, marking the entry of agentic computing into aerospace.
UK AI Minister Kanishka Narayan announced plans to strengthen cyber defenses, rejecting reliance on tech giants' self-regulation and urging frontier AI firms to adopt proactive safety measures. The government also emphasized building sovereign AI infrastructure to address rapidly escalating AI capabilities in cybersecurity.
Interconnects reports a shift in open-source model releases from dominance by a few Chinese players to a globally diversified ecosystem. New entrants like Zyphra, Cohere, and Poolside highlight the rise of 'pure' model makers and sovereign AI initiatives. This trend reflects expanding technical breadth and strategic adjustments amid geopolitical pressures.
OpenAI has launched the 'OpenAI for Australia' initiative to build sovereign AI infrastructure and accelerate local innovation. The program aims to upskill over 1.5 million workers to support the country's growing AI ecosystem.
OpenAI announced a new agreement with the UK Ministry of Justice to enable civil servant access to ChatGPT and offer UK data residency options for its API, Enterprise, and Edu platforms. This move supports the earlier MOU goals of advancing sovereign AI capabilities and public sector adoption in the UK.
NVIDIA released Nemotron-Personas-India to support sovereign AI via synthetic data. However, the provided text contains only the title, lacking technical details, scale, or evaluation results to verify its actual contribution.