KUALA LUMPUR: Artificial intelligence (AI)-powered scanners at the country's main entry points are not yet fully effective in identifying goods accurately due to insufficient data, says Datuk Hasbi Habibollah.
The Deputy Transport Minister said the system was still in its learning stage, with more data being collected to improve its detection capabilities.
He said this was observed during visits to locations where such scanners had been installed, including Port Klang and the Malaysia Airlines cargo area.
“The use of high-tech scanners, including those equipped with AI, in cargo inspections has the potential to enhance detection capabilities through image analysis and the identification of objects with specific characteristics.
“However, the effectiveness of any system depends on the type of technology, system configuration, database, algorithms, detection accuracy, as well as the ability and competency of the security inspectors operating the system.
“When it comes to import and export operations at airports, we do have scanners equipped with AI features. However, to be frank, the system is still in its learning stage.
“I have seen for myself that when the AI generates an image, it is still unable to identify the items shown accurately. This means its data is not yet sufficient.
“However, officers on the ground are currently collecting data to feed into the AI system. This is to improve the effectiveness of AI detection, as recommended by the Malaysian Anti-Corruption Commission (MACC), and help our officers carry out their duties more effectively,” he said during the question and answer session in the Dewan Rakyat on Monday (Nov 2).
Hasbi was responding to a supplementary question from Kuala Kangsar MP Datuk Iskandar Dzulkarnain Abdul Khalid on the current capabilities of AI-powered scanners at cargo terminals to detect the smuggling of dangerous goods.
He added that the ministry was also considering working with other countries to share data and strengthen the capabilities of the AI system.
He said this approach would not only save time but also facilitate the collection of data needed to train the system.
He stressed that AI technology should be regarded as a supporting tool in the security inspection process, rather than the sole basis for deciding whether cargo should be cleared.
Inspection decisions should be based on the overall established security control process, including image assessments by security inspectors, the use of approved inspection methods, further inspections where necessary and compliance with approved security programmes, he added.
“In this context, AI can help enhance inspectors’ ability to identify anomalies or objects requiring further attention.
“However, inspection decisions and follow-up actions must remain subject to approved security procedures and verification by qualified personnel.
“The government will continue to assess the suitability and effectiveness of the inspection technologies in use based on technological developments, threat levels, risk assessments and current supply chain security requirements,” he said.
