Governments want the productivity of AI agents without giving up control of their data. On November 5, 2026, Cohere announced a strategic partnership with Singapore’s Centre for Strategic Infocomm Technologies (CSIT) that puts that trade-off to the test in one of the world’s most security-conscious environments: national security operations. CSIT will deploy Cohere’s enterprise-grade AI models and its agentic AI platform, North, inside highly secure environments — a move that follows a successful pilot and signals how sovereign AI is moving from policy papers into real operations.
What the Partnership Covers
Through the partnership, CSIT will leverage Cohere’s enterprise-grade AI models and North to accelerate the deployment of AI applications in highly secure environments, in support of Singapore’s national security operations. The collaboration includes building and evaluating potential applications including secure knowledge retrieval, automated analysis and report generation, and agentic workflows designed to streamline national security operations. CSIT is an agency of Singapore’s Ministry of Defence and a member of the Defence Technology Community. It builds advanced cybersecurity and secure digital solutions, conducts applied research in cybersecurity, and develops AI and data exploitation tools in a self-built secure cloud environment — in its own words, its engineers’ code protects a nation, counters digital threats, and secures its most sensitive secrets. Cohere, founded in 2019, positions itself as a provider of sovereign AI for governments and regulated industries: AI systems that organizations own, operate and govern, with control over their data, infrastructure and deployment. The deal follows a successful pilot in which Cohere demonstrated it could deploy and operate AI inside secure on-premises environments while integrating with CSIT’s existing systems and workflows. Andrew Chang, Cohere’s Vice President of APAC, said the company is “deeply committed to helping enterprises and governments build long-term sovereign AI capabilities.”
North: The Agentic Platform at the Center
North is Cohere’s agentic AI platform: a ready-to-use workplace application that chats over an organization’s internal documents, grounds answers in connected data, creates and refines documents, and builds customizable agents and workflow automations. Crucially for CSIT, North can be deployed and run in an organization’s own environment — on-premises or fully air-gapped — with strong security, data and compliance controls. Its North Admin panel lets teams enforce guardrails at the individual agent level, and the platform is model-agnostic, so customers can use Cohere’s models or bring their own. The secure knowledge retrieval use case at the heart of the CSIT deal maps directly onto North’s retrieval strengths; our guide to retrieval-augmented generation explains the underlying pattern of grounding a model in authoritative internal data before it answers. A similar enterprise retrieval idea is behind Box’s Nemotron-powered retrieval engine for enterprise AI agents, but the CSIT deployment adds a constraint Box’s customers rarely face: the data must never leave a classified boundary. Cohere’s recent North 2 upgrade, launched about a month ago, unified enterprise-grade security, intelligence, cost governance and full-stack control into a single platform — the same pitch CSIT is now buying into.
Why Sovereign AI Matters for National Security
Sovereign AI is the answer to a problem every government IT buyer recognizes: the most capable AI models are usually delivered as cloud APIs, but the most sensitive government data cannot legally or practically leave the building. The result has been a two-track market — public-cloud AI for ordinary work, and heavily guarded on-premises systems for everything else. Partnerships like this one try to close that gap by bringing the AI inside the perimeter instead. The CSIT deal is part of a broader APAC push for Cohere, which also works with Fujitsu, LG CNS and Japan’s Digital Agency on secure, enterprise-grade AI. It also continues a direction CSIT has pursued for years: in November 2023 it began piloting Google Distributed Cloud Hosted in CSIT’s earlier sovereign cloud pilot with Google Cloud, and the Cohere deal extends that same logic from infrastructure to the AI layer itself. The defense-grade security mindset shows up across the industry right now — from post-quantum security in embedded hardware to OpenAI’s GPT-6 Cyber security model, which finds and helps patch software vulnerabilities. The common thread is AI moving from a productivity tool into the security stack itself.
What Comes Next — and What Is Still Unproven
What the announcement does not say matters as much as what it does. No contract value, timeline, or number of users has been disclosed, and the applications named — secure knowledge retrieval, automated analysis and report generation, agentic workflows — are described as potential applications to be built and evaluated, not deployed systems. The pilot is the only published evidence that the platform works inside CSIT’s environment so far, and the release offers no independent evaluation or benchmark results. That is normal for a government technology partnership at this stage, but it means the real test is still ahead: whether North’s agents can handle classified workflows at operational scale without leaking data, hallucinating in reports, or creating new attack surface inside a secure network. Read the Cohere and CSIT announcement and Cohere’s North platform pages, and CSIT’s official site, for the full details. If the rollout succeeds, it becomes a reference case other governments will study — proof that agentic AI can work inside the most guarded networks, not just alongside them.
The Cohere–CSIT partnership is a small announcement with a large signal: sovereign AI is no longer just a procurement preference, it is becoming an operational capability inside national security. The pilot proved the platform can run on-premises; the partnership will test whether agentic AI can earn trust where trust matters most. Watch what CSIT actually deploys in the coming months — that, more than any press release, will show whether sovereign AI has arrived.
