AI & Automation

AI Model Inference Is Now Available In-Region in Singapore: What It Means for Regulated Data

November 13, 2026 · 5 min read · Etzal Group Team

On 14 October 2026 Anthropic announced the opening of an office in Singapore, its fifth in Asia Pacific, and the signing of a memorandum of understanding with Singapore's Ministry of Digital Development and Information (MDDI). The same announcement said that in-region inference via Amazon Bedrock is now available in Singapore for regulated sectors.

The news is a month old, and the source is the company's own release. We are covering it for one practical reason: on 16 January 2027 Indonesia's rules on cross-border transfers of personal data take effect, and many companies are deciding where their AI processing is allowed to run. A sentence about the in-region AI inference Singapore now offers is easily read more broadly than it means.

What was announced

According to Anthropic's announcement:

  • The Singapore office is the company's fifth in the Asia Pacific region.
  • The memorandum with MDDI was signed on the sidelines of Singapore International Cyber Week by Kai Fong Chng, Permanent Secretary of MDDI, and Ben Mann, a co-founder of Anthropic.
  • It covers cooperation on AI model safety, security and evaluation, research into AI adoption and its effects, and responsible adoption in the public and private sectors, including support for small and mid-sized businesses adopting AI.
  • In-region inference via Amazon Bedrock is now available in Singapore for regulated sectors.

The announcement also carries several figures, all of them company claims. In Anthropic's Economic Index, Singapore ranks second of 121 countries for Claude.ai use relative to population, nearly six times what its size would predict. More than 1,500 engineers and technical staff at GovTech Singapore are said to use Claude Code, and Grab's sales assistant BriX is said to save its sales teams about 40,000 hours each quarter.

What the source does not explain: which sectors count as "regulated", and what technical conditions apply. Ask the provider for the detail before building it into a design.

Inference location is not storage location

This is where most misunderstanding starts. There are several different locations, and an announcement like this usually speaks about one.

The questionWhat it means
Inference locationWhere the model processes your request
Storage locationWhere inputs, outputs and logs are kept, if they are kept
Access locationWhere provider staff can view data from, for support or monitoring
Your own data locationWhere your source databases and documents sit

In-region inference answers the first row. The second and third must be asked separately. Processing can happen in Singapore while logs are kept elsewhere, or the reverse. We are not saying that is the case for this service, since the source gives no detail. We are saying you need to ask.

Personal data from Indonesia

This part should be stated without hedging: for personal data originating in Indonesia, inference in Singapore still means the data is sent abroad. Singapore is nearer and in the same region, but it is another country.

According to Veritask's summary of Government Regulation (Peraturan Pemerintah, or PP) 33/2026, transfers of personal data abroad follow an order of legal bases set out in Article 165, and controllers must map their transfer flows and assess the risk. We take that article number from a third-party summary, not from the official text, so check it against the official regulation before citing it.

In practice there are three different situations:

  • A Singapore company with Singapore customer data. In-region inference can help meet demands from a sector regulator or from customers that processing stay in the country. Confirm your sector's requirements with the regulator concerned.
  • An Indonesian company sending personal data to a model running in Singapore. This is a cross-border transfer. The Singapore location does not remove the duty to map flows and assess risk.
  • A group with entities in both countries. You may need two different answers for two data sets.

Obligations differ from company to company. Confirm your own case with legal counsel or the data protection authority.

Questions for the provider

This list applies to any model provider, not only the one in this news.

  1. In which region does inference run, and are there circumstances where requests are routed elsewhere, for example under heavy load?
  2. Are inputs and outputs stored? Where, for how long, and for what purpose?
  3. Can retention be switched off or shortened, and on what conditions?
  4. Who at the provider can access the data, and from which country?
  5. Is the data used to train models?
  6. Are all models available in that region, or only some?
  7. What documents can you give us as evidence for our auditors or regulators?

The first and sixth questions are about trade-offs. Restricting processing to one region can mean fewer model choices, tighter capacity at peak load, or different pricing. That is a general possibility, not something stated in the source. Ask before you commit.

Mapping your AI data flows

The duty to map transfer flows, as summarised by Veritask, can begin with a simple worksheet. One row per AI use:

  1. The business process and its purpose.
  2. The type of data entering the model: any personal data, and whose.
  3. Origin: the country where the data was collected.
  4. Inference location.
  5. Storage location and duration at the provider.
  6. The transfer basis relied on, filled in with legal counsel.
  7. Reduction measures: masking, dropping fields, or sending no personal data at all.

Column seven is often the cheapest way out. Many AI tasks do not need names, identity numbers or phone numbers. Removing them before the data is sent shrinks the transfer problem, wherever inference runs.

The decision order we suggest

  1. Start with the data, not the vendor. What does the model actually need to see?
  2. Reduce first. Strip personal data the task does not need.
  3. Decide the acceptable location for what remains, with legal counsel.
  4. Then choose a provider and region that meets it, with written answers to the seven questions above.
  5. Record the decision. The flow map and risk assessment are the documents you will be asked for.

If you would like that flow map drawn up and the architecture options compared, Etzal Group's AI consulting team can help.

Sources

Frequently asked questions

What does in-region inference mean?

The AI model processes requests in data centres inside a given territory, here Singapore, instead of routing them to another region.

Does inference in Singapore settle obligations for personal data from Indonesia?

No. Sending personal data from Indonesia to Singapore is still a transfer abroad and still follows the transfer rules that apply.

What should I ask a provider besides inference location?

Where inputs and outputs are stored, for how long, who can access them, and whether any other processing happens outside the region.

Author

Etzal Group Team

Etzal Group Team, helping businesses grow with technology. Contact us.

Related service

WhatsApp AI Chatbot

We build, train, and run a WhatsApp AI agent that answers customers 24/7, qualifies leads, and follows up.