|
Anthropic has finally given its Singapore customers a focal point. The San Francisco‑based AI lab opened an office in the city‑state on Wednesday—its fifth in the Asia‑Pacific—and signed a memorandum of understanding (MOU) with the Ministry of Digital Development and Information (MDDI) that places AI safety testing at the heart of the partnership.
The MOU was signed on Tuesday, on the sidelines of Singapore International Cyber Week, by MDDI Permanent Secretary Chng Kai Fong and Anthropic co‑founder Ben Mann, with Minister for Digital Development and Information Josephine Teo and Anthropic’s Head of APAC Policy Michael Sellitto looking on. It covers three areas: AI safety, security and model evaluation; research on how AI is adopted and what it does to the economy; and responsible adoption across government and industry.
The agreement obliges Anthropic to engage in technical exchanges with Singapore’s AI safety, security and assurance bodies over model evaluation, testing standards, and incident‑response procedures—the playbook for managing misbehaviour in production. Additionally, Anthropic will contribute research and data from its Anthropic Economic Index, assist small‑ and medium‑sized enterprises in adopting AI tools, develop AI capabilities among public officials, and provide policy advice as Singapore refines its AI regulations.
This cooperation is not unprecedented. Since joining the Generative AI Evaluation Sandbox launched jointly by IMDA and the AI Verify Foundation in 2023, Anthropic participated in red‑teaming efforts across English, Mandarin, Malay, and Tamil to uncover model vulnerabilities. The new MOU codifies a partnership that was already in progress.
Key details remain absent. The MOU does not disclose investment volumes, postings such as staffing targets, or the financial backing of the SME and skill‑building initiatives. As typical for MOUs, these documents express intent rather than enforcing contractual obligations, a practice Singapore frequently employs.
What is actually in the deal
The agreement obliges Anthropic to engage in technical exchanges with Singapore’s AI safety, security and assurance bodies over model evaluation, testing standards, and incident‑response procedures—the playbook for managing misbehaviour in production. Additionally, Anthropic will contribute research and data from its Anthropic Economic Index, assist small‑ and medium‑sized enterprises in adopting AI tools, develop AI capabilities among public officials, and provide policy advice as Singapore refines its AI regulations.
This cooperation is not unprecedented. Since joining the Generative AI Evaluation Sandbox launched jointly by IMDA and the AI Verify Foundation in 2023, Anthropic participated in red‑teaming efforts across English, Mandarin, Malay, and Tamil to uncover model vulnerabilities. The new MOU codifies a partnership that was already in progress.
Anticipating substantial demand, the primary rationale for establishing a local presence lies in its established customer base. Overseas talent exceeds 1,500 engineers and technical staff from the Government Technology Agency of Singapore utilise Claude Code, its AI‑driven development assistant, to modernise public sector systems. Meanwhile, the Cyber Security Agency explores leveraging Claude to bolster national cyber‑defence capabilities.
Within the finance sector, major players already integrate Cloud Code: DBS employs it for software development and operational efficiency, whereas sovereign wealth fund GIC has deployed Claude across its investment and corporate service units. Practically regulated entities can now operate Claude via in‑region inference on Amazon Bedrock, ensuring data stays local—a consideration heightened by the Monetary Authority of Singapore’s forthcoming AI risk rulemaking clock.
The leading tech conglomerates have also turned to Anthropic. Grab employs its Claude‑powered sales assistant, BriX, delivering roughly 40,000 hours of saved effort each quarter. Sea leverages Claude to accelerate its Shopee customer‑service chatbot, making interactions quicker and more accurate, while thousands of employees write and debug code using the model. Founder Supabase, a Singapore‑origin developer platform, is further integrating Claude Code into its ecosystem.
A notable dependence risk surfaces. In June, Anthropic temporarily halted access to its Fable and Mythos models for roughly three weeks to comply with U.S. Department of Commerce export restrictions, later restoring availability on 1 July. This episode underscores that ultimate control over frontier‑grade models remains with Washington. The MOU makes no mention of such regulatory exposure, prompting queries about how GovTech would manage a scenario in which U.S. sanctions affect core model availability.
The commercial dimension presents another challenge. Anticipating broader client acquisition, the office also serves as a growth engine; however, detailed financial commitments to SMEs and skill‑building programmes remain unspecified. While MOUs clarify intent, they do not guarantee substantive resource allocation for those initiatives.
The customer list is the real story
Anticipating substantial demand, the primary justification for expanding the footprint centers on the deep engagement already present. More than 1,500 engineers and technical staff from the Government Technology Agency of Singapore utilise Claude Code, its AI‑driven development assistant, to modernise public sector systems. Meanwhile, the Cyber Security Agency explores leveraging Claude to strengthen national cybersecurity defences.
Industry partners are equally active. In finance, DBS leans on Anthropic’s Cloud Code for rapid software development and efficiencies, while the sovereign wealth fund GIC has rolled out Claude across investment and corporate service functions. Regulated firms now run Claude via in‑region inference on Amazon Bedrock, keeping data within the jurisdiction—a factor amplified as the Monetary Authority of Singapore moves forward with AI risk regulation timelines.
A key caveat highlights market concentration. Anthropic’s investor disclosures reveal two customers generating nearly a quarter of its 2025 revenue; the company’s future compute spend tops US$518 billion. Scaling beyond a narrow set of large U.S. clients is essential for diversification, and the Singapore office therefore functions both as a safety alliance and a revenue outlet.
Rivals are circling the same prize
The competitive environment reinforces this positioning. OpenAI recently appointed ex‑Elastic and Snowflake veteran Sanjay Deshmukh to lead Asia‑Pacific sales, and has backed a public‑private AI accelerator in Thailand. Google advances its strategy through school partnerships, telecom investments, and localized language rollouts for Gemini. Databricks has announced a US$350 million capital infusion for the city‑state, while the indigenous SEA‑LION model offers heritage‑language solutions for Southeast Asia.
Distinctively, Anthropic emphasizes safety evaluation above consumer reach—an area where Singapore’s AI Verify unit has earned a reputational edge. For a laboratory whose reputation hinges on coupling capability with rigorous safety assessment, a government actively assessing models stands as a natural collaborator.
The questions an MOU cannot answer
Two persistent challenges cast doubt on the deal’s completeness. First, dependency: Export constraints imposed by the United States—most visibly manifesting in the temporary restriction of Fable and Mythos access during June—show that external regimes, not the host state, ultimately determine model availability. The MOU provides no recourse regarding such regulatory threats.
Second, commercial urgency: The strategic imperative to broaden client portfolios runs alongside clear financial signals about the importance of growing beyond a limited set of major U.S. accounts. While the MOU outlines potential contributions, explicit budgets for SME onboarding and workforce skill‑development remain undefined, leaving implementation risks unresolved.
Despite these points, the arrangement remains a logical step for Singapore. Access to advanced evaluation techniques fortifies the island nation’s claim to be Asia’s AI assurance hub. Moreover, a community exceeding 300 engineers and researchers—attending the latest Builder Summit—now enjoys a direct channel to Anthropic expertise. Going forward, the success of the partnership will be measured by concrete outcomes: published testing standards, actionable incident‑response protocols, and well‑funded SME initiatives. Until those achievements materialise, Singapore retains a powerful new resident and an added layer of institutional detail.
