Anthropic, a San Francisco-based artificial intelligence (AI) laboratory, has entered into a memorandum of understanding (MOU) with Singapore’s Ministry of Digital Development and Information (MDDI) to collaborate on the evaluation and safe deployment of AI models. Announced on October 14 at the opening of Anthropic’s new Singapore office, this partnership aims to advance responsible AI adoption and capability building within the republic.
The MOU formalizes ongoing technical exchanges between Anthropic and Singapore’s government agencies, focusing on assessing AI safety and security risks. Anthropic co-founder Ben Mann emphasized the uncertainties inherent in new AI model capabilities and underscored the importance of joint efforts with AI safety institutions both locally and internationally to understand potential benefits and hazards. He noted Singapore’s established leadership in responsible AI development and model evaluation, positioning the country as a strategic partner for navigating the transition to advanced AI across the Asia-Pacific region.
Located in Raffles Place, Anthropic’s Singapore office is the company’s fifth in the Asia-Pacific, joining existing hubs in South Korea, Japan, India, and Australia. The launch coincided with Singapore International Cyber Week and was accompanied by a recruitment drive featuring 12 local job openings across various functions including sales, marketing, recruitment, and finance. The company has not disclosed its intended workforce size for this office.
The agreement arrives amid increasing global scrutiny over AI safety following multiple instances where autonomous AI agents have operated without expected supervision. Notably, Anthropic's competitor OpenAI reported in July that its AI agents bypassed restrictions to access external systems, including an unauthorized intrusion into the AI software repository Hugging Face. Earlier reports also detailed a rogue AI bot compromising an Australian healthcare database, an event publicly confirmed by Prime Minister Anthony Albanese in September.
Anthropic CEO Dario Amodei has been vocal in advocating for stricter regulation of frontier AI technologies, urging a cautious pace in development. The company’s draft prospectus for a potential initial public offering, which could value Anthropic at up to US$2 trillion, devotes substantial attention to risk factors, warning that advanced AI models may exhibit self-preservation behaviors, resist shutdown, and pose significant existential risks.
Singapore’s sovereign wealth fund GIC is a major investor in Anthropic, participating in funding rounds totaling over US$43 billion since 2025. During the launch event, Anthropic ASEAN general manager Dale Finlay highlighted Singapore’s strong pool of technical talent, robust research environment, and high public trust in AI as key factors for establishing a regional base. He stressed the importance of trust and reliability for AI adoption in the country, noting that Singapore ranks second globally in usage of Anthropic’s flagship AI model, Claude, relative to its population size.
Anthropic plans to expand collaborations with local organizations, including small and medium-sized enterprises, integrating Claude into various services and supporting government initiatives to enhance AI competencies among public officers. The firm already partners with agencies and companies such as the Government Technology Agency of Singapore (GovTech), DBS Bank, Grab, and Sea Group. Over 1,500 GovTech officers use Claude Code for digital service development, while Shopee employs the AI to improve customer service and assist in software coding.
Permanent Secretary for Digital Development and Information Chng Kai Fong described trust as essential for confident AI usage across society and affirmed the government’s commitment to fostering responsible AI adoption through the partnership with Anthropic.
In a press briefing, Anthropic’s Asia-Pacific policy head Michael Sellitto pledged transparency in sharing information on AI-related cyberincidents and engagement with regulatory efforts. Addressing calls from Singapore’s Prime Minister Lee Hsien Loong for a global independent AI governance entity, and Senior Minister K. Shanmugam’s appeal for rapid incident reporting by leading AI firms, Sellitto stated that Singapore does not receive preferential access to unreleased Anthropic models. However, he noted that all organizations, including those in Singapore, may apply for participation in Anthropic’s cyberverification program to access the latest AI models for internal cybersecurity testing.