Latest news with #AISociety
The Star
5 days ago
- The Star
‘Collection of metadata poses risks'
PETALING JAYA: Like puzzle pieces scattered across a table, bits of digital data may appear meaningless on their own. But with enough time, as well as location and behavioural clues, a recognisable picture can emerge. That is the concern raised by cybersecurity experts over a government initiative to collect anonymised mobile phone data. The Mobile Phone Data (MPD) programme, introduced by the Malaysian Communications and Multimedia Commission (MCMC), is intended to support public policy, particularly in tourism and infrastructure planning. Although authorities have emphasised that the data excludes names and identification numbers, experts warn that by combining the anonymous data with other metadata such as tower location, timestamps and user behaviour, it could still expose individuals to reidentification and cyber threats. According to AI Society president Dr Azree Shahrel Ahmad Nazri, even coarse location data such as cell tower logs can be used to build a person's detailed behavioural profile. 'From just a few days of movement data, researchers can predict who you are with over 90% accuracy,' he claimed when contacted. 'This is why metadata is not truly anonymous.' MCMC, in a media briefing last week, clarified that IMEI numbers and SIM card IDs were not among the data fields requested. However, Azree Shahrel cautioned that even without those identifiers, centralising metadata still poses significant cybersecurity risks. He also warned that such repositories could become high-value targets for hackers, cybercriminals or foreign actors. 'If breached, this data could form a detailed map of user routines, enabling highly targeted attacks or surveillance,' he said. He suggested that persistent identifiers, such as anonymised mobile numbers, be replaced with session-based tags, and that precise timestamps be aggregated to reduce the risk of tracking individuals. Universiti Malaysia Sarawak lecturer Chuah Kee Man echoed those concerns, pointing out that the MPD does not currently violate the Personal Data Protection Act 2010 (PDPA), as anonymised metadata and government agencies fall outside its scope. However, he argued that this legal blind spot still raises red flags. 'The collection is occurring without the public's explicit consent or even knowledge. 'And while it may not breach the PDPA directly, it creates ethical and legal issues surrounding the erosion of privacy rights,' he said. He warned that once data is stored at this scale, it could potentially be used for political profiling, social control or surveillance. 'The integrity of how this data is used relies entirely on those managing it – both now and in the future,' he said. He called for a shift in approach, including the principle of data minimisation, where only essential data is collected, and for the implementation of informed consent policies. 'If the government insists on collecting such data, it must demonstrate a clear need and adopt every possible measure to protect users,' he said. Cybersecurity specialist Fong Choong Fook said public concern about the MPD programme is not unfounded, especially given previous data breaches involving government-linked agencies. 'One of the most notable cases was in 2017, when the personal data of 46 million Malaysians was leaked after the MCMC outsourced work to a contractor. 'Incidents like these continue to shape public scepticism,' he said. The massive data breach in 2017, believed to affect almost the entire population of Malaysia, included lists of mobile phone numbers, identity card numbers, home addresses and SIM card data of 46.2 million customers from multiple mobile phone and mobile virtual network operators. 'Take note that the PDPA does not apply to MCMC. This means that if a data leak were to occur, MCMC would not be held liable,' he said, highlighting a gap in accountability. Fong urged the government to be transparent about the anonymisation process and to release a clear set of guidelines outlining how the data is managed, what safeguards are in place and how privacy is protected. 'There should be a publicly accessible framework, or at least a white paper that can be scrutinised by independent experts. 'We cannot continue operating in a black box,' he said.
The Star
03-06-2025
- Business
- The Star
Made-in-Malaysia AI vital for growth
Experts: Time to build, own and export our own systems so as not to be left behind PETALING JAYA: Malaysia must urgently develop its sovereign AI systems or risk becoming digitally advanced but economically dependent – and ultimately face long-term economic decline, experts say. They said as Malaysia races ahead with its digital transformation, much of that progress remains powered by imported artificial intelligence tools, thus raising concerns about economic dependency. AI Society president Dr Azree Shahrel Ahmad Nazri said Malaysia must urgently develop and own its AI systems or risk falling further behind economically, despite the country's rapid digital transformation. 'Without local AI ownership, we cannot export, licence or embed these technologies into regional products that generate income or foreign exchange,' he said when contacted. He warned that countries like Malaysia, which rely heavily on foreign-developed AI, face growing risks to GDP growth, productivity and national competitiveness. Citing emerging macroeconomic trends, he explained how AI-driven productivity gains, if limited to non-tradable sectors like education, health or governance, can distort real growth indicators. In Malaysia's case, he said the impact could be severe. Despite being online, automated and connected, he said the country risks building a digital economy that 'exports little, owns less and pays more', driven by rising licensing fees, AI subscriptions and reliance on foreign cloud infrastructure. Azree said such imported productivity could also distort Malaysia's exchange rate, making the ringgit appear stronger than it really is and hurt traditional export sectors. CLICK TO ENLARGE He called for urgent investment in sovereign AI capacity and policy-driven infrastructure. 'We must build, own and eventually export our own AI. Otherwise, we risk being tenants in our own digital economy,' he said. Universiti Malaysia Sarawak senior lecturer Dr Chuah Kee Man said Malaysia's dependence on foreign AI tools may bring short-term efficiency, but it risks long-term loss of control and competitiveness. 'To become an AI exporter, Malaysia must move beyond short-term grants and build a genuine innovation ecosystem,' he said. This includes regulatory sandboxes, stronger intellectual property protection and targeted investment in areas like halal logistics, Islamic finance, smart cities and multilingual natural language processing. Chuah explained that Malaysia's current progress is uneven across sectors. While the country has talent and early-stage initiatives, he said sovereign AI development demands sustained investment in infrastructure, skills and computing power. In terms of cost, he said developing a medium-sized foundational model would require serious commitment. 'If we take current industry standards for medium-sized foundational models, it could be about RM50mil and take about two years to build if there is a consolidated effort on it. 'For a sovereign AI model, it could easily go up to RM500mil. It can take up to five years or longer if the infrastructure is not sufficiently powerful,' he said. Because of the scale involved, Chuah suggested a collaborative approach. 'Heavy investment – that is why a partnership model may be more feasible since we are technically playing catch-up,' he said. Echoing the need for strategic investment, cybersecurity specialist Fong Choong Fook said Malaysia should establish a National AI Sovereignty Fund to finance homegrown AI development. Fong said the sovereignty fund can be backed by capital from public and government-linked companies. 'This fund would directly support the development of models like PutraGPT, prioritising Bahasa Melayu/Nusantara, Islamic jurisprudence and regional languages,' he said. Fong said developing the nation's own AI system would also pave the way for a Malaysian Large Language Model (LLM) consortium, open data-driven AI innovation and export-ready frameworks. 'The LLM could bring together universities, MNCs, GLCs and start-ups to train sovereign models using ethically-sourced local datasets. 'Government agencies should anonymise and release datasets (health, law, commerce) to stimulate trustworthy AI development when adopting open data for AI innovation. 'We can also develop export frameworks and certifications – such as halal certification standards – so that AI models are certifiable for compliance, ethics and privacy, especially to build trust in Islamic and Asean markets,' he added. When developing AI, Fong said the government must consider investing in three strategic pillars: talent, infrastructure and IP ownership. He said Malaysia needs a strong pipeline of AI engineers, data scientists and computational linguists, supported through focused education, public-private partnerships and incentives to retain talent locally. He added that Malaysia must also establish national AI compute infrastructure, including sovereign cloud services, GPUs and secure data centres to support large-scale model training and experimentation. 'IP ownership is essential to shift policies to prioritise ownership of algorithms, datasets and models – especially for sectors like finance, halal trade, cybersecurity and Bahasa Melayu/Nusantara applications where we have domain edge and regional relevance.' 'If we can develop AI that is context-aware for Asean or Islamic financial systems, Malaysia can lead in culturally and linguistically aligned AI exports, rather than competing head-on with Silicon Valley,' he said. With partnerships like MyEG-BeiTou, where Malaysia contributes data and policy access while the core technology remains foreign-owned, Fong said the country must remain cautious when relying on foreign-based AI systems. 'While it may represent short-term digital progress by enabling adoption and government integration, it also sets a precedent: data flows out, while value flows into others. This is digital dependence masked as digital development. 'True progress means co-developing and co-owning the intellectual property. If Malaysians provide the data, the insights and the use case validation, then Malaysians must also hold equity in the resulting tech,' he added.
