Anthropic Opens Singapore Office and Partners with Government

Anthropic opened its fifth office in the Asia-Pacific region in Singapore on Wednesday (14), signing a memorandum of understanding with the local government focused on AI safety and responsible adoption. The document was signed by Kai Fong Chng, Permanent Secretary of the Ministry of Digital Development and Information, and Ben Mann, co-founder of Anthropic, at the Singapore International Cyber Week with Minister Josephine Teo and Michael Sellitto, the company's head of public policy for the region, as witnesses.

The memorandum addresses three areas: model safety and assessment, research on AI adoption and its effects, and responsible use in public and private sectors. According to TNGlobal, Anthropic envisions technical exchanges with Singapore's AI safety and assurance institutions regarding model assessment, testing standards, and incident response.

Clients Prior to the Office

The company arrives with an established client base. Over 1,500 engineers and technicians from GovTech, Singapore's government technology agency, utilize Claude Code to modernize public systems, according to Anthropic. At Grab, the sales assistant BriX, built on Claude, saves approximately 40,000 hours per quarter for sales teams, also according to the company. DBS employs the model for software development and productivity, and the sovereign wealth fund GIC is implementing it in investment and corporate service teams.

These figures come from Anthropic itself and have not been audited by third parties. The company also did not disclose investment amounts or the number of employees for the new office, a notable omission in specialized coverage. The relationship with the government, however, is longstanding: since 2023, Anthropic has participated in the Generative AI Evaluation Sandbox, led by IMDA and the AI Verify Foundation.

Two Regulators, Two Approaches, on the Same Day

While Ben Mann signed a technical cooperation agreement in Singapore, the European Commission announced in Dublin formal information requests to Anthropic and OpenAI based on the AI Act regarding incidents where models from both labs reached external systems during tests. The same company received, within hours, an invitation to jointly assess models and a regulatory questionnaire.

It would be easy to interpret this as Singapore being permissive versus Brussels being stringent, but that reading would be incorrect. The Singapore memorandum addresses precisely model assessment, testing standards, and incident response—the same themes the Commission wants documented. The difference lies in the instrument: Singapore focuses on negotiated technical access before the problem; the EU uses regulation afterward. Both pathways create obligations for the lab, but only one comes with a response deadline and a possibility of sanctions.

What Singapore's Choice Signals for the Region

Singapore serves as a regional hub for banks, asset managers, and integrators serving Southeast Asia, and the names mentioned by Anthropic (DBS, GIC, Grab) affirm the financial and platform profile of the client base. The channel to other markets is the same that brought the company here: cybersecurity in the financial sector.

In Japan, MUFG Bank, Sumitomo Mitsui Banking Corp., and Mizuho Bank negotiated access to Claude Mythos in May, according to Bloomberg and Nikkei, and the Ministry of Finance established a public-private working group on the cyber risk of these models. The Singapore memorandum formalizes with a government what Japan has organized through banks.

In India, the vector is service provision. HCLTech announced on the 12th a partnership with Anthropic to bring Claude, Cowork, and Claude Code to corporate clients, coinciding with a report of advanced AI revenue of $219 million for the quarter. Indian integrators operating from Singapore for Southeast Asian clients now have, in the same city, both the model provider and the regulator assessing it.

The absence of a figure in the announcement has an effect that extends beyond Singapore. Governments in the region negotiating similar agreements with American labs will have a precedent of a technical cooperation memorandum centered on assessment and incident response, not a commitment to investment. For CIOs of Asian banks, this means that the question of who tested the model, and with what access, is likely to arrive before the question of how much the provider will invest in the country.