Building the PETRONAS of data

EVERY time we subscribe to an artificial intelligence (AI) tool or digital service, we may unknowingly be buying back our own data.

Whether you are a small and medium enterprise (SME) or a government agency, few stop to consider that many of today’s AI models and digital products have been trained, in part, on our collective digital footprints.

This lies at the heart of the modern AI race: a new form of data colonialism, where global technology firms harvest user data – often in exchange for “free” services – refine it into high-value AI products, and ultimately sell those products back to the very people and countries from which the data originated.

For developing nations, protecting data from exploitation via data privacy laws and governance is no longer enough.

We must actively transform data – particularly government-held data (information generated, collected, or managed by public sector agencies) – into a strategic economic commodity.

However, achieving this at scale remains a hurdle; while governments sit on massive troves of raw information, unlocking its value requires specialised machine learning refinement skills that remain scarce in the public sector.

Sovereign data advantage

The global data market is approaching a tipping point that presents an opportunity for governments. Frontier AI models have largely exhausted the supply of high-quality, human-generated data available on the public Internet.

To continue improving their models, many AI developers are increasingly relying on synthetic data – content generated by AI rather than humans.

This triggers the risk of a “synthetic data loop” – a phenomenon where AI models train on AI-generated content, causing errors and biases to compound over generations until the model suffers total performance collapse.

As a result, authentic, human-verified data has become a strategic asset.

Governments hold a unique advantage as trusted custodians of decades of real-world, population-scale data – from healthcare and education to taxation and socioeconomic records – that continuously updates.

These datasets provide a critical safeguard against the limitations of synthetic data.

However, like crude oil, raw data has little value until it is cleaned, structured, secured, and refined into a usable asset.

Monetisation gap

Malaysia is not starting from zero. Over the past two years, the Digital Ministry has established several key foundations for a secure digital economy.

The Data Sharing Act 2025 provides the legal framework for inter-agency data sharing, complemented by strengthened protections under the Personal Data Protection Act.

At the same time, the National AI Office is driving the country’s AI agenda; the proposed Sovereign AI Cloud aims to strengthen digital sovereignty, and plans to establish a Malaysia Data Commission reflect a maturing data governance ecosystem.

These are important milestones, but they primarily strengthen governance and protection. They establish the rules for securing and sharing data, not for creating economic value from it.

Defensive policies can prevent misuse and theft – but on their own, they do not transform Malaysia’s data assets into a source of national wealth.

Escaping data colonialism

To monetise government data at scale, Malaysia needs a bold new strategy.

Today, valuable public-sector data is fragmented across agencies, while limited specialist talent and bureaucratic constraints make commercialisation difficult.

One must remember that data only creates significant value when it is aggregated, refined, and responsibly managed.

To succeed, we need a more unified approach.

We should create a commercial state-owned enterprise (SOE) – a specialised, business-minded organisation similar to how Petroliam Nasional Bhd (PETRONAS) manages the nation’s oil and gas – to act as the official steward for the country’s “digital oil.”

Consolidating this function under a single SOE would allow the country to pool scarce expertise to monetise Malaysian data.

Instead of asking civil servants in separate agencies to figure out the path forward, this new PETRONAS of data with specialised talents in business and tech would be empowered to deploy sophisticated monetisation strategies – such as partnerships with data companies in the ecosystem, tiered premium application programming interfaces, national data marketplaces and secure data sandbox leases – while maintaining data protection.

There are currently five different types of players who are monetising data in different ways in the data refinement ecosystem:

> Asset owners – consolidate raw data, for example Reddit, Meta, TikTok and Shutterstock.

> Intermediaries – manage secure access and licensing, for example TollBit, Human Native and AWS Data Exchange.

> Data refiners – clean and structure data, for example Scale AI, Appen and TELUS Digital.

> Infrastructure providers – provide secure storage and processing vaults, for example Databricks, Snowflake and Kleene.ai).

> AI developers – transform refined data into AI models and digital products, for example OpenAI, Google and Anthropic.

The SOE’s path to commercialisation will likely evolve over three phases modelling the different player types:

> Phase 1: The niche data asset owner. Catalogue high-value, niche proprietary public data pools that global tech giants cannot replicate, such as tropical agricultural telemetry, localised multilingual datasets, Islamic finance transaction trends, and more.

> Phase 2: The data refiner and vault. Refine these raw niches into machine-ready assets.

Depending on the data privacy sensitivity, certain data can be securely licensed, while highly sensitive assets are permanently retained within national data vaults. Beyond revenue, the SOE can then strategically trade these different data points with global AI firms for infrastructure access and technology transfer.

> Phase 3: Nationwide scaling and niche sovereign AI models. Expand capabilities to translate any public or regulated data pool into premium, highly secure commercial commodities.

SOE can eventually build its own domain-specific AI models, develop foundational sovereign language models if we so desired.

If executed correctly, this SOE will evolve into a global powerhouse.

Much like PETRONAS transformed from a local manager of oil into a global energy leader, the SOE can eventually have subsidiaries globally that are exporting our data-sovereignty blueprint to help other nations reclaim their digital wealth.

In the AI era, the phrase “data is the new oil” is not a cliché – it is an economic reality.

Malaysia must act decisively now, or risk continued data colonisation by global tech giants.