Anthropic opened an office in Singapore on Wednesday, the company's fifth in the region, and signed an agreement with the Ministry of Digital Development and Information on AI safety, security and model testing.
Co-founder Ben Mann signed the memorandum with Chng Kai Fong, the ministry's permanent secretary, during the Singapore International Cyber Week conference. Digital Development and Information Minister Josephine Teo and Michael Sellitto, who leads Anthropic's APAC policy work, attended the signing, the company said in its announcement.
The document covers safety and security work, evaluation of models, studies of how AI is taken up and what effects it has, and responsible use by government bodies and companies. Anthropic said it plans technical exchanges with Singapore's safety and assurance institutions on testing standards and incident response, and will share data from the Anthropic Economic Index.
"When we train a new model, nobody really knows exactly what its capabilities are," Mann said at the Anthropic Builder Summit at Capella Singapore, according to CNA. He said working with safety institutes such as the Singapore AI Safety Institute was important to understanding the risks and benefits of new models.
Chng said the partnership reflected the government's commitment to scaling responsible adoption of AI and deepening capabilities in companies and agencies.
Neither the office nor the memorandum came with a published figure or headcount. Anthropic said the office is open and hiring, without giving a target. The company said Singapore comes second among 121 countries for Claude.ai use per head of population, about six times higher than the country's size would suggest. That figure comes from its own Economic Index, which measures use of its products rather than AI adoption overall.
More than 1,500 engineers and technical staff at GovTech Singapore use Claude Code, according to Anthropic. Grab's sales assistant BriX, built on Claude, saves its sales teams about 40,000 hours a quarter, and DBS and sovereign wealth fund GIC are rolling out Claude. The country's cyber security agency is also looking at ways Claude might strengthen its defences.
Singapore has signed similar agreements with other AI companies this year. It reached a deal with OpenAI in May, and the Infocomm Media Development Authority signed a memorandum with Microsoft in June on AI safety and security. Anthropic joined the agency's Generative AI Evaluation Sandbox in 2023 and later took part in a red-teaming exercise covering four of the country's languages.
The Next Web noted that Singapore is not covered by US export controls that name entities and places, and that Anthropic does not allow Chinese firms or their overseas subsidiaries to use its most advanced models.