“Guidelines needed”:
East Asian countries are under huge diplomatic pressure to choose between the EU and the US for their AI regulatory framework, officials said.
-
Sherry Shan / Staff Reporter
The Ministry of Digital Affairs (MODA) announced yesterday that it will closely monitor the evolution of agent-based artificial intelligence (AI) before making decisions about its integration into government systems.
Digital Minister Lin Yi-jing made the remarks during a meeting of the Legislative Transport Committee when asked about security issues related to OpenClaw and other agent AI applications.
Open-source AI agents have come under scrutiny after their use was found to be vulnerable to cybersecurity attacks, allowing malicious attackers to hack machines, steal credentials, and gain access to sensitive data.
Photo: Tu Chien-jung, Taipei Times
China restricts the use of OpenClaw by government agencies and companies and warns about it in its cybersecurity manual.
Nvidia CEO Jensen Huang (黃仁勳) announced Monday at a technology conference in San Jose, Calif., that the company is introducing NemoClaw, which adds privacy and security controls to AI agents to make them more “reliable, scalable, and accessible” for users around the world.
Taiwan People’s Party (KMT) Legislative Committee Member Jackie Chan ( Chen Qinglong ) and Chinese Nationalist Party (Kuomintang) Legislative Committee Member Huang Jianhao (黃凃洗) asked Mr. Lin how the government plans to regulate the use of AI agents and whether there are any plans to integrate AI agents into the operation of government systems.
Unlike ChatGPT, developed by OpenAI, agent AI applications such as OpenClaw automatically help users upload videos to YouTube or book airline tickets, Lin said.
“People can converse in ChatGPT just as they would in the real world. However, using OpenClaw would raise cybersecurity concerns, as some users have reported that AI agents automatically made purchases,” Lin said.
He said the US, China and the EU have different views on AI development.
While the European Union is wary of the technology and has published a list of prohibited applications under its AI law, the U.S. government has been reluctant to impose limits on AI, believing regulations would stifle innovation, Lin said.
In China, he said, whatever the Chinese government says about AI applies.
Singapore, Taiwan, Japan and South Korea are under significant diplomatic pressure to choose between the EU and US regulatory frameworks, Lin said.
“We want to find a balance between these two models,” Lin added. “Personally, I believe that in the long term, agent AI will be integrated into the operation of government systems to improve efficiency. However, because this technology is so new, there may be cybersecurity risks that have not yet been identified.”
He said the government should propose regulatory guidelines for AI agents after monitoring their continued evolution, adding that they may not necessarily be the same as those proposed by big technology companies.
The national AI governance framework began with the AI Basic Law (Artificial Intelligence Basic Law), followed by the AI Risk Classification Framework, which government agencies use to handle health, financial and transportation data, Lin said.
Huang Chien-hao asked whether the government’s plan to spend NT$10 billion (US$314.1 million) over 10 years on AI development was appropriate, given that some other countries are spending more on AI-related programs.
Lin said it is important for Taiwan to increase its computing power for AI.
He said projects to build AI computing power centers in other countries are mostly initiated and funded by private investors.
In Taiwan, AI computing centers are classified as critical basic infrastructure and will be built using government and private funds, Lin said.
“We will discuss this with the Financial Supervisory Commission.” [FSC] “We are allowing investment from private insurance companies. So far, Hon Hai Precision Industry has applied to invest in an AI computing center,” Lin said.
Separately, members of Congress’ Finance Committee expressed concerns about the risks of using OpenClaw for stock trading, citing the potential for errors in stock quantities.
FSC Chairman Peng Jinlong said the commission had issued guidelines on the application of artificial intelligence in the financial industry, but further research was needed on the issue.
He did not say when guidelines for such AI agents would be published.
