Anthropic opens Singapore office and signs government AI safety agreement

Anthropic opened its Singapore office on 14 October and announced a memorandum of understanding (MOU) with Singapore’s Ministry of Digital Development and Information (MDDI) covering AI safety, model evaluation and adoption. Anthropic co-founder Benjamin Mann and MDDI Permanent Secretary Chng Kai Fong signed the agreement on 13 October, on the sidelines of Singapore International Cyber Week, CNA reported. Minister for Digital Development and Information Josephine Teo and Anthropic’s head of Asia-Pacific policy, Michael Sellitto, witnessed the signing.

MDDI agreement sets out technical exchanges

Under the MOU, Anthropic says it plans to exchange technical information with Singapore’s AI safety, security and assurance institutions on how models are evaluated, how they are tested against standards and how incidents are handled. The agreement also covers research on the adoption and impact of AI, alongside responsible use across government and business. Anthropic says it intends to share research and data, including material from its Anthropic Economic Index, and offer technical and policy expertise as Singapore develops AI rules.

The distinction between exchanging information and obtaining access to a model matters to the work covered by the agreement. Sellitto said the MOU does not give Singapore priority access to Anthropic’s frontier models before their commercial release, The Business Times reported. It does cover testing and evaluation. Sellitto also said Anthropic was committed to informing governments about AI-related incidents of concern “as quickly as we responsibly can”.

Singapore’s Senior Minister and Coordinating Minister for National Security, K Shanmugam, said on 13 October that the government wanted “full transparency immediately” and faster information-sharing when AI risks arise, CNA reported. He directed that request towards major technology companies with offices in Singapore. The MOU’s planned exchanges on incident response sit alongside that request, while the two statements put different words around the timing of information-sharing.

Anthropic joined Singapore’s evaluation sandbox in 2023

Anthropic says its work with Singapore’s AI institutions began in 2023, when it joined the Generative AI Evaluation Sandbox run by the Infocomm Media Development Authority (IMDA) and the AI Verify Foundation. It subsequently worked with them on a red-teaming project in English, Mandarin, Malay and Tamil. The planned exchanges under the new MOU extend a relationship that has already involved both model evaluation and testing in Singapore’s languages.

Sellitto said Anthropic had held a session with Singapore’s AI Safety Institute to train government and civil society experts in evaluating models and assessing risks, The Business Times reported. Mann described safety evaluation as an important part of the new arrangement. “When we train a new model, nobody really knows exactly what its capabilities are,” he said at Anthropic’s Builder Summit, according to CNA. He said work with safety institutes could help establish a clearer understanding of new models’ risks and benefits as they are developed.

The Anthropic agreement is one of several involving AI developers and Singapore’s government. Singapore signed a deal with OpenAI in May concerning applied AI innovation, and IMDA signed an MOU with Microsoft in June on AI safety and security. Singapore has also formed a cyber task force and launched an agentic AI testing kit, as AI Affairs previously reported.

Raffles Place office and onshore models

Anthropic’s new office is at Raffles Place. It is the company’s fifth in Asia-Pacific, after offices in Bengaluru, Seoul, Sydney and Tokyo. Anthropic had announced the Singapore opening in September. AI Affairs has previously reported on OpenAI and Anthropic expanding their Singapore offices as premium rents climbed.

At the office launch, Mann said Anthropic was working to bring its frontier models “on soil” in Singapore, but had “no specific dates to share on that”, The Business Times reported. Anthropic separately says organisations in regulated sectors can run Claude with in-region inference on Amazon Bedrock. Its work to bring frontier models onshore is a further plan, with its timing left open by Mann’s account.

Anthropic is building its Singapore team around sales and customer work. Dale Finlay, its general manager for ASEAN, said the immediate emphasis was on developing those operations for customers in the public sector, enterprises and startups. The company declined to give its current headcount or hiring target. Chris Ciauri, Anthropic’s managing director of international, said the company had no immediate plan for a core research operation in Singapore, although that could change as its research teams assessed local talent.

GovTech’s Claude Code adoption

Anthropic’s own Economic Index ranks Singapore second among 121 countries for Claude.ai usage relative to population, at nearly six times the level its population alone would predict. That is a company measure of use of its own service, rather than a measure of AI adoption across Singapore. Anthropic says more than 1,500 engineers and technical staff at the Government Technology Agency of Singapore (GovTech Singapore) have adopted Claude Code for work on government systems and digital services.

The company also names DBS, GIC, Grab and Sea among Singapore organisations using Claude in their operations. It says it will look for ways to help small and medium-sized businesses adopt AI responsibly and to support AI capability among public officers. Anthropic says the Cyber Security Agency of Singapore is exploring whether Claude could help strengthen the country’s cyber defence.