Anthropic’s expansion in Singapore began before its new office doors opened. In 2023, the company joined the country’s Generative AI Evaluation Sandbox and later worked with local institutions on red-teaming Claude in English, Mandarin, Malay and Tamil.¹

That groundwork culminated this week at Singapore International Cyber Week, where Anthropic co-founder Ben Mann and Kai Fong Chng, permanent secretary at the Ministry of Digital Development and Information, signed a memorandum of understanding. The agreement covers AI safety, security and model evaluation, research into adoption and its effects, and responsible use across government and business.¹

For Singapore, the attraction is practical as well as regulatory. Chng said the state’s approach has been to enable innovation while keeping up with changing risks, arguing that trust is essential if companies, workers and the public are to use AI confidently.¹ Anthropic, meanwhile, points to a market already deeply embedded in its products: more than 1,500 GovTech engineers and technical staff use Claude Code, while DBS, GIC, Grab and Sea are among customers deploying its tools.¹

But the launch also exposed the limits Anthropic places on local control. Asked about data residency and sovereign-AI arrangements, Mann said the company wants customers to run their most sensitive work with it, yet must preserve the ability to detect misuse because the technology is “dual use.”²

That stance puts Anthropic at odds with the appeal of open-weight models, including Chinese alternatives drawing users with lower prices and flexibility. Mann warned that once such models are public, “they can’t be put back in the box.”² The company says it is working to bring frontier models onto Singapore-based infrastructure, but offered no timetable—leaving the city-state to balance access, autonomy and the guardrails Anthropic refuses to surrender.²