At Anthropic’s Builder Summit in Singapore on Wednesday, Oct. 14, co-founder Benjamin Mann said the company would retain oversight of how customers use its most powerful AI models. He cited the technology’s “dual use” nature: Anthropic wants customers to run sensitive workloads on its systems while detecting misuse. Mann was in Singapore for the opening of Anthropic’s fifth Asia-Pacific office.

In September, Anthropic introduced its Enterprise Frontier Safeguards program, which offers stronger data privacy protections, including zero data retention for eligible customers, while retaining safeguards to detect misuse. Anthropic routinely reports attempts by state and non-state actors to use Claude for military or other harmful purposes. One alleged case involved a weapons development cell in northern Yemen seeking to use the model for ballistic missile development.

According to company data, Singapore ranks second among 121 places in Claude use relative to population, at nearly six times the level its population size would predict. Anthropic says more than 1,500 engineers at GovTech use Claude Code; DBS, GIC and Grab also work with Claude. OpenAI opened a Singapore office in late 2024 and committed 300 million Singapore dollars ($234 million) to the local AI ecosystem. Singapore has also signaled interest in Chinese open-source models; its homegrown SEA-LION is built on Alibaba’s Qwen.

Open-source models from Chinese labs including DeepSeek, Moonshot AI and z.ai are gaining users through low prices, near-frontier performance and flexibility. Demand for Chinese models overtook U.S. models on OpenRouter this year. Anthropic international managing director Chris Ciauri said the company does not see AI as a zero-sum game. Mann said open-weight models cannot retain safeguards resilient to being released publicly, warning that countries encouraging that ecosystem share a broader risk.

Anthropic CEO Dario Amodei has urged AI companies to slow releases by a year or two so oversight and security work can catch up. In June, U.S. export restrictions on Anthropic’s Fable 5 and Mythos 5 models led the company to suspend worldwide access; the restrictions were lifted less than three weeks later. Anthropic’s APAC policy head Michael Sellitto said democratic, rights-respecting countries should lead frontier AI development and deployment. Mann said Anthropic was working as quickly as possible to bring its frontier models onto infrastructure in Singapore, without giving dates.

At Anthropic’s Builder Summit in Singapore on Wednesday, Oct. 14, co-founder Benjamin Mann said the company would retain oversight of how customers use its most powerful AI models. He cited the technology’s “dual use” nature: Anthropic wants customers to run sensitive workloads on its systems while detecting misuse. Mann was in Singapore for the opening of Anthropic’s fifth Asia-Pacific office.

In September, Anthropic introduced its Enterprise Frontier Safeguards program, which offers stronger data privacy protections, including zero data retention for eligible customers, while retaining safeguards to detect misuse. Anthropic routinely reports attempts by state and non-state actors to use Claude for military or other harmful purposes. One alleged case involved a weapons development cell in northern Yemen seeking to use the model for ballistic missile development.

According to company data, Singapore ranks second among 121 places in Claude use relative to population, at nearly six times the level its population size would predict. Anthropic says more than 1,500 engineers at GovTech use Claude Code; DBS, GIC and Grab also work with Claude. OpenAI opened a Singapore office in late 2024 and committed 300 million Singapore dollars ($234 million) to the local AI ecosystem. Singapore has also signaled interest in Chinese open-source models; its homegrown SEA-LION is built on Alibaba’s Qwen.

Open-source models from Chinese labs including DeepSeek, Moonshot AI and z.ai are gaining users through low prices, near-frontier performance and flexibility. Demand for Chinese models overtook U.S. models on OpenRouter this year. Anthropic international managing director Chris Ciauri said the company does not see AI as a zero-sum game. Mann said open-weight models cannot retain safeguards resilient to being released publicly, warning that countries encouraging that ecosystem share a broader risk.

Anthropic CEO Dario Amodei has urged AI companies to slow releases by a year or two so oversight and security work can catch up. In June, U.S. export restrictions on Anthropic’s Fable 5 and Mythos 5 models led the company to suspend worldwide access; the restrictions were lifted less than three weeks later. Anthropic’s APAC policy head Michael Sellitto said democratic, rights-respecting countries should lead frontier AI development and deployment. Mann said Anthropic was working as quickly as possible to bring its frontier models onto infrastructure in Singapore, without giving dates.