Anthropic signed a memorandum of understanding with Singapore’s Ministry of Digital Development and Information on 14 October, the same day it opened a Singapore office. The document sets out where the two sides intend to cooperate. It is not a purchase, and the announcement describes no contract value, no deployment and no obligation on either party.
That distinction is the whole story, so it is worth being exact. Permanent Secretary Kai Fong Chng signed for the ministry and co-founder Ben Mann for Anthropic, at Singapore International Cyber Week. Minister Josephine Teo and Michael Sellitto, Anthropic’s head of Asia-Pacific policy, witnessed it. Three areas are named: responsible adoption across government and industry, studies of how AI spreads and what it changes, and joint work on AI safety, plus how models are tested and secured.
Almost every commitment in Anthropic’s own post is phrased as a plan. On the safety side, Anthropic says it plans exchanges between its engineers and Singapore’s AI assurance institutions, covering incident response and testing standards. It would also lend technical and policy advice while Singapore drafts its AI rules, and hand over research and data (including from its Economic Index). It also plans to help small businesses adopt AI and to build AI skills among public officers. None of this carries a date or a target.
The relationship predates the paper. Anthropic joined a government-backed sandbox for testing generative AI in 2023, run by the Infocomm Media Development Authority and the AI Verify Foundation, and later worked with them on a red-teaming exercise, meaning deliberate attempts to break a model, in English, Mandarin, Malay and Tamil. The MOU formalizes a track record rather than starting one.
The office is an expansion. Anthropic calls it its fifth in Asia-Pacific. It says the office is open and hiring but gives no headcount, no hiring target and no breakdown of roles. Dale Finlay, general manager for ASEAN, is the only named local executive. The commercial case in the post rests on existing customers: more than 1,500 engineers at the government technology agency GovTech Singapore use Claude Code, the bank DBS and the sovereign wealth fund GIC use Claude, and Grab says a Claude-powered sales assistant saves roughly 40,000 hours a quarter. Those figures are Anthropic’s, and the post does not say whether Grab’s number comes from Grab. The Cyber Security Agency is described only as “exploring” a use. Singapore also ranks second of 121 countries in Claude.ai usage relative to population, according to Anthropic’s own index.
Our read, not the source’s: Singapore is a sensible place to anchor a frontier lab’s Southeast Asia work. It is a regulatory and financial hub with a state that tests AI systems itself, and a safety-and-evaluation agreement gives Anthropic a seat while rules are drafted. Other labs court the same governments, and the post says nothing about rivals or about competing agreements, so the competitive reading is ours.
The practical effect is local support and a policy channel, not a new government contract. For companies in regulated Singapore sectors, the item with a concrete date is that Claude now runs with in-region inference on Amazon Bedrock. Whether the MOU produces anything beyond meetings will show in the first published joint evaluation or the first shared dataset, and none has been promised yet.
Based on Anthropic’s newsroom post of 14 October 2026.
