Regulations essential for foreign AI like China’s DeepSeek, says expert

SL Rajesh of the IACSP says this is crucial to ensure data privacy and prevent foreign influence permeating Malaysia, amid concerns of biased content and censorship.

Last year, digital minister Gobind Singh Deo said Putrajaya was giving serious consideration to China’s DeepSeek and studying how it can be adapted for local use to keep pace with rapid AI developments.

PETALING JAYA:

 A cybersecurity expert has urged Putrajaya to draw up clear rules to regulate foreign artificial intelligence (AI) platforms like China’s DeepSeek, to ensure data privacy and curb foreign influence in Malaysia.

SL Rajesh of the International Association for Counterterrorism and Security Professionals Centre (IACSP) said AI engines are not neutral in nature but shaped by the nation from which they originate.

He said this meant AI platforms may reflect their local laws, biases as well as political views.

Rajesh, who heads IACSP’s computer forensics department, added that lower‑cost AI systems such as DeepSeek have been found to disseminate inaccurate or biased information in response to user queries.

“For Malaysia, this is crucial. If AI systems like DeepSeek influence areas like policymaking, education or public communication, they might unintentionally shape public opinion in ways that don’t align with Malaysia’s values.

“Misinformation can cause harm — whether it’s influencing elections, spreading false health information, or creating social tension. If DeepSeek mismanages sensitive topics, it could make misinformation worse,” he told FMT.

Developed by Chinese researchers and engineers, DeepSeek made waves early last year and was touted as a potential rival to ChatGPT, particularly due to its specialisation in Mandarin and other Asian languages.

It also stood out because developers could adapt the platform to their specific use while processing more data per second at a much lower cost than the competition.

However, several nations, including Korea, Germany, Australia, Taiwan, and Czech Republic, have banned the use of DeepSeek over data security concerns, particularly in government departments.

Apart from DeepSeek, Taiwan’s national security bureau has also flagged other AI models such as Doubao, by TikTok’s parent company ByteDance Ltd, and Baidu’s Ernie Bot, citing biased content and censorship.

Gobind Singh Deo.

In February last year, digital minister Gobind Singh Deo said Putrajaya was giving “serious consideration” to China’s DeepSeek and studying how it can be adapted for local use to keep pace with rapid AI developments.

In November, Gobind said his ministry expects to present an AI legislative framework to the Cabinet in June 2026.

Flat out ban not the way to go

Rajesh said data security concerns arise when information entered into these platforms is processed and stored abroad. This could include sensitive government and national intelligence information, as well as the personal data of citizens.

Nonetheless, he argued that an outright ban on the technology would be unwise.

Instead, he urged the establishment of clear rules to govern what data may leave Malaysia and the implementation of risk assessments on foreign platforms before approving their use.

“The government should create clear rules for foreign AI companies like DeepSeek. These rules should ensure providers explain how their AI platform works and makes decisions.

“It should also ensure that the AI is fair and doesn’t show bias or make mistakes that could harm people, and that they go through regular checks to ensure it follows Malaysian laws.”

Need for Malaysian-made AI

Farlina Said of Malaysia’s Institute of Strategic and International Studies said these concerns justify the need for Malaysian-made AI to safeguard the data and infrastructure that support its functions.

Farlina Said.

“The control is necessary for compliance with Malaysian laws and the regional standards Malaysia subscribes to. It would also ensure the AI is fit-for-purpose, thus utilising the necessary resources needed to achieve the aims of the company or ministry,” she said.

Rajesh agreed, saying this would allow the technology to meet Malaysia’s cultural and national needs while boosting security, supporting ethical standards and reducing dependence on foreign platforms.

Farlina said a good rule of thumb for the government and companies is to take a cautious approach to foreign AI models. While DeepSeek is cost effective, she said users would expect to place trust in its gatekeeping capabilities.

Nonetheless, she pointed out that a Stanford University report found that the privacy practices of other AI models were not necessarily better, necessitating greater internal guidance on privacy practices in Malaysia.

“All models do have their merits, Concerns about potential misinformation are also relevant for all models. It is best practice to verify sources, and if possible, to seek out the exact document the AI had yielded the information from.” - FMT