Safeguards pushed as AI agents gain autonomy

An OpenAI test incident underscores risks from systems with less human supervision. 

Singapore will put safeguards in place as artificial intelligence (AI) agents become capable of taking more actions with less human supervision, according to Prime Minister Lawrence Wong.

“As technology becomes more powerful, we also have to guard against risks that are harder to predict,” he told the National Day Rally on 23 August.

The push comes as Wong cited an OpenAI agent that went beyond its test environment and broke into another company’s systems during testing. 

“No one told the AI agent to attack the company. But in trying to complete its tasks, the agent did things which its developers did not want it to do,” he noted.

The government will also seek to work with other countries and companies on safeguards for the technology.

‘Stricter age checks’

Meanwhile, Singapore will require social media platforms to introduce robust and reliable age checks to enforce minimum age requirements for young users.

“Self-declaration is just not good enough,” Wong said.

The government will also require stronger safeguards for teenagers aged 13 and above, as it seeks to address platform features designed to keep users scrolling, clicking, and returning.

“If its safeguards remain inadequate, we will consider additional restrictions – including raising the minimum age for access to that platform beyond 13,” he added.

The measures come as one in six people aged 10 to 24 in Singapore shows signs of problematic social media use. Wong said for some, usage can reach six or seven hours a day, sometimes extending into the early hours of the morning.