The Price of Cheap AI: ASEAN's Switching Question

AI companies are engaged in a price war, which is a good deal for Southeast Asian businesses, at least for now. Top AI models cost only about US$20 a month, and even the free versions are still more than sufficient. However, this situation seems too good to be true. As a matter of fact, Anthropic and OpenAI are the two most valuable private companies in the world today, each approaching a trillion dollars, but neither is profitable, not even in the near term.

Low prices but no profit usually mean AI companies are competing to capture market share. Initially, low prices attract new customers. As these customers become more dependent on one provider, they build layer after layer of integration. This increasing interconnection makes them 'stickier.' Once their overall workflow, data, and staff are thoroughly embedded with a particular provider, switching to another becomes too expensive. At that point, staying with the original provider is often the only viable option.

Where ASEAN Sits

There are concerns that ASEAN is beginning to move in this direction. The Singaporean government has secured its first MoU with OpenAI for research, talent, and public service applications. In 2024, Grab announced a strategic collaboration with OpenAI and later built merchant-facing agentic tools with both OpenAI and Anthropic. GoTo's GoPay launched Dira, an Indonesian-language voice assistant, built on Gemini 1.5 Flash. While these developments may suggest that AI adoption in ASEAN is still in its early stages, the potential long-term risks of lock-in should not be underestimated.

Data portability, interoperability, and clear switching terms that keep low prices a product of competition rather than a lure into dependence on one provider are important policy tools to build now, while dependence is still shallow. Regulators need to design them in a way that does not compromise intellectual property, cybersecurity, or commercial confidentiality.

Data Portability and Interoperability

AI has made portability and interoperability far more complex than they used to be. Effective portability policies must distinguish between individual and enterprise users, as well as between raw and processed data.

For individuals, portability focuses on moving prompts, files, chat history, preferences, and saved outputs between AI assistants. In contrast, enterprises may need to transfer greater configurations and larger amounts of institutional knowledge from a specific system.

The same differentiation applies to data. Portability regulations usually protect raw data, such as documents or customer records. However, AI lock-in often comes from processed or inferred data, such as recommendations and behavioural insights. Even the EU Data Act restricts its portability scope to raw and pre-processed data only, arguing that inferred or derived data falls under intellectual property rights protection.

Therefore, portability policies must be designed with careful consideration to avoid disrupting the market.

How ASEAN Should Move Forward

Research from ERIA on data governance indicates that among member states with comprehensive privacy laws, the only universally recognized data-subject right is the right of access. There is no regional consensus on data portability rights. Existing national portability provisions are limited in four ways: they do not cover non-personal, machine-generated data; they lack an interoperability mandate; they do not address cloud-switching to prevent vendor lock-in; and they remain scoped to individual data subjects, with no equivalent right for enterprise or business users. These gaps are crucial to address for fostering competition, especially as AI adoption grows, a point echoed in an ERIA brief on competition in the digital economy, which noted that an ex-ante approach is increasingly relevant to prevent anti-competitive behaviour. The good news is that DEFA offers a platform to align this work across the region, as highlighted in subsequent ERIA research.

DEFA negotiations concluded in late May 2026. Public text of the agreement is not yet available, and it remains unclear whether data portability was addressed directly in the negotiated provisions on data governance and privacy. If it was not, implementation and future protocol work under DEFA still gives ASEAN room to build a standard that covers machine-generated data alongside personal data, mandates interoperability rather than conditioning portability on it, extends switching obligations to cloud providers, and creates a parallel enterprise-inclusive portability right.

AI labs will continue to compete, but for ASEAN, it is essential to ensure that users have meaningful ways to switch between services and benefit from this competition. A competitive AI market should focus on value rather than dependency. With DEFA's substantive terms now largely settled, the more realistic opening for portability and switching rights lies in implementation and future protocol work, where the region can still shape how these commitments are operationalized before AI dependencies deepen further.

This opinion piece was written by was written by Fikri Adib, Research Associate at ERIA, Ibrahim Naufal, Research Associate at ERIA, and Jordan Holbrook, Digital Innovation and Sustainable Economy Economist at ERIA. It was published in Borneo Bulletin.

Disclaimer: The views expressed are solely those of the author and should not, under any circumstances, be regarded as representing the official position of the Economic Research Institute for ASEAN and East Asia.