Artificial intelligence company Anthropic opened a Singapore office on Tuesday and signed a memorandum of understanding with the Ministry of Digital Development and Information covering AI safety, security and model evaluation, the firm said.
Permanent Secretary Chng Kai Fong and Anthropic co-founder Ben Mann signed the pact on the sidelines of Singapore International Cyber Week, with Minister Josephine Teo and Anthropic’s head of APAC policy witnessing the ceremony, according to the company’s statement.
Why Singapore matters to Anthropic
Anthropic described Singapore as one of its most active markets, citing public-sector and commercial use of its Claude models by organisations including GovTech, Grab and DBS. The local office is the company’s fifth in Asia-Pacific and will focus on customer engineering and policy engagement rather than training frontier models from scratch.
Grab told investors its Claude-powered sales assistant saves about 40,000 staff hours each quarter, while Sea Group uses Claude for Shopee customer service and internal coding tools. Those case studies matter to regulators weighing how generative AI sits inside critical digital infrastructure.
Regulatory alignment
The MOU threads into Singapore’s wider push on trusted AI: sandbox deployments, vendor liability for chatbots at banks, and the newly passed Digital Infrastructure Act licensing large data centres. Anthropic’s safety research agenda overlaps with CSA guidance on securing model supply chains.
MDDI has emphasised that adoption must pair with evaluation tooling and incident reporting. Signing with a frontier lab gives Singapore a channel to stress-test policies before they harden into binding codes.
Labour and skills
Anthropic said it is hiring locally for policy, solutions and partnerships roles. The move competes with other model providers expanding in the region, but differs in positioning around constitutional AI and third-party testing.
For technology workers, the office signals demand for engineers who can integrate APIs under enterprise security rules — less hype, more compliance-aware deployment. Enterprises still bear liability when vendor models hallucinate; MAS has made that clear for banks.
What comes next
The MOU is not a procurement contract. It sets up research exchanges and responsible adoption pilots. Watch for joint publications on red-teaming and for references in IMDA’s next AI governance consultation.
Singapore’s bet remains unchanged: attract frontier labs, then bind them to local safety norms faster than markets alone would.
Competitive field
OpenAI, Google and local telco-backed clouds already operate enterprise AI programmes in Singapore. Anthropic’s differentiation rests on safety research and Claude’s context window, but procurement still runs through IMDA-aligned security reviews for government workloads.
Startups building on Claude will need to document data residency and prompt logging under PDPA guidance. The new office gives them a local escalation path when models update weekly.
Enterprise adoption
Chief information officers surveyed at a Wednesday DBS forum said entry-level automation risk is real, but they still plan pilots in compliance-heavy functions first — exactly the use cases Anthropic markets with audit trails.