Anthropic opens its Singapore office and signs an AI safety MOU with the government

What did Anthropic announce in Singapore?

Anthropic said on 14 October 2026 that it was expanding its Asia-Pacific operations with a new office in Singapore, its fifth in the region, and signing a memorandum of understanding (MOU) with the Singapore Ministry of Digital Development and Information. The company called Singapore “one of our most active markets anywhere in the world.” The same text was also published by the Singapore Economic Development Board, labelled as a press release issued by Anthropic.

The office was formally launched at an event at Capella Singapore on the sidelines of Singapore International Cyber Week, according to The Straits Times, which reported that the office is in Raffles Place and that 12 local roles, in sales, marketing, recruitment and finance, were listed on Anthropic’s careers site that day. Anthropic declined to give a planned headcount. The Business Times noted that the opening had first been announced in September and that the other Asia-Pacific offices are in Bengaluru, Seoul, Sydney and Tokyo.

Dale Finlay, Anthropic’s general manager for ASEAN, framed the move around reliability: “Our customers here hold AI to an extremely high standard, especially when it comes to reliability and security. Having a team on the ground means we can get even closer to the teams putting Claude to work on their hardest problems.” Anthropic also ran a Builder Summit for more than 300 developers on the day, after a community meetup earlier in the week that it says drew more than 400 local builders.

What does the MOU with MDDI actually cover?

The MOU was signed by Chng Kai Fong, Permanent Secretary of MDDI, and Anthropic co-founder Ben Mann, with Minister for Digital Development and Information Josephine Teo and Michael Sellitto, Anthropic’s head of Asia-Pacific policy, as witnesses. Anthropic lists three areas of cooperation: AI safety, security and model evaluation; research on AI adoption and its impact; and responsible adoption across the public and private sectors.

The most concrete commitment is technical exchange. Anthropic says it plans to work with Singapore’s AI safety, security and assurance institutions on model evaluation, testing standards and incident response, the same kind of work it does with safety institutes elsewhere. The company’s argument is that sharing technical information gives governments a clearer view of what frontier models can do and what risks they carry, while helping developers build safer systems.

Beyond testing, Anthropic says it will look for ways to build AI skills in Singapore, help small and medium-sized businesses adopt AI responsibly, and support the government’s goal of building AI capability among public officers. It also plans to share research and data with the government, including from the Anthropic Economic Index, and to offer technical and policy expertise as Singapore develops its AI rules.

Mann put the testing rationale bluntly at the launch, as quoted by The Straits Times: “When we train a new model, nobody really knows what exactly the capabilities are. Working with AI safety institutes in Singapore, as well as those around the world, is important to understand what the risks and the benefits could be.” Chng said the partnership reflects the government’s commitment to scaling up responsible adoption and that trust is critical for businesses, workers and the public to use AI with confidence.

The relationship is not new. Anthropic says it joined the Generative AI Evaluation Sandbox run by the Infocomm Media Development Authority (IMDA) and the AI Verify Foundation in 2023, and later worked with them on a red-teaming project in English, Mandarin, Malay and Tamil. The MOU formalises and widens that work rather than starting it.

How is Claude already used in Singapore?

Anthropic backs its “most active markets” claim with its own usage data. In its latest Economic Index, Singapore ranks second of 121 countries in Claude.ai usage relative to population, at nearly six times what its size alone would predict.

In government, Anthropic says more than 1,500 engineers and technical staff at the Government Technology Agency (GovTech) have adopted Claude Code to modernise systems and speed up software delivery, and that the Cyber Security Agency of Singapore is exploring how Claude could help strengthen national cyber defence. In finance, it names DBS, which uses Claude for software development and productivity in its Singapore operations, and GIC, the sovereign wealth fund, which it says is rolling Claude out across investment and corporate services teams. Regulated customers can run Claude with in-region inference on Amazon Bedrock.

Among technology companies, Anthropic says Grab’s BriX sales assistant, built on Claude, saves Grab’s sales teams about 40,000 hours a quarter; that Sea uses Claude to make Shopee’s customer-service chatbot faster and more accurate, with thousands of its employees coding with it; and that Supabase is building with Claude Code. These are vendor-reported figures; the customers’ own numbers were not published alongside them.

GIC is also an investor. The Straits Times reported that the fund first invested in Anthropic in September 2025 and led a US$30 billion Series G round in February 2026, a reminder that the commercial and policy relationships here overlap.

What the MOU does not include

Anthropic was careful about what the agreement does not do. Its executives told reporters that the MOU does not give Singapore priority access to unreleased frontier models ahead of commercial release, though it does cover testing and evaluation work. The Straits Times quoted Sellitto as saying any Singapore organisation can instead apply to Anthropic’s Cyber Verification Program to use the latest released models for internal security testing.

Nor is there a date for hosting the newest models locally. “We are working as quickly as we can to bring the frontier models on soil in Singapore, but no specific dates to share on that,” Mann told The Business Times, which also reported that Anthropic has no plans for a research operation in Singapore for now. No funding, binding obligations, evaluation protocols or timelines were published with the MOU.

On incidents, Anthropic made a softer promise. Reports quote its executives as committing to share information about AI-related cyber incidents with governments “as quickly as we responsibly can.” The Straits Times linked that to recent calls from Prime Minister Lawrence Wong for a global independent body to govern AI and from Senior Minister K. Shanmugam for frontier labs to share incident information promptly.

Why Singapore matters to frontier AI labs

The timing is deliberate. The signing happened during a Cyber Week in which Singapore had already refreshed its national cybersecurity strategy with AI-specific measures, and in which Google DeepMind and GovTech showed an on-device Gemma scam detector for Android. Singapore is positioning itself as a place where frontier labs test, deploy and report, not just sell.

For Anthropic, an MOU with a mid-sized government that has years of model-testing experience is a template it can reuse. It resembles the work Anthropic says it already does with safety institutes elsewhere, and lands as governments push labs for more transparency, from the UK AI Security Institute’s eight technical priorities to the European Commission’s information requests on loss-of-control incidents. For Singapore, it fits a strategy built on testing and evaluation rather than heavy regulation.

What is still unclear

Several things are not yet public:

• The MOU text itself has not been published; its scope comes from Anthropic’s announcement and press reporting, and MOUs of this kind are generally not binding.
• Which Singapore institutions will run the technical exchanges, how often, and whether any evaluation results will be published was not stated.
• Usage and productivity figures, including the Economic Index ranking, the 1,500 GovTech users and Grab’s 40,000 hours, are Anthropic’s own.
• There is no timeline for in-country hosting of frontier Claude models or for local headcount beyond the roles listed on launch day.