Singapore backs international call for greater frontier AI safeguards

SINGAPORE: Singapore is backing an international call from more than 20 world leaders for stronger safeguards for frontier artificial intelligence (AI) amid greater scrutiny of the technology’s risks.

Worries about frontier AI resurfaced in recent weeks after leading scientists and top lab executives warned that its development could outpace abilities to control increasingly capable models. Some even warned of the risk of possible human extinction if AI development does not slow down.

Prime Minister Lawrence Wong said on Tuesday (Sep 22) that Singapore is supporting a joint statement on AI safety initiated by Norwegian Prime Minister Jonas Gahr Støre and Finnish President Alexander Stubb.

The letter, which is still open for endorsement, has been signed by Mr Wong as well as 21 other world leaders, including Canadian Prime Minister Mark Carney and Australian Prime Minister Anthony Albanese.

It called for companies to develop transparent safety protocols, including mandatory pre-deployment testing and independent evaluation, with qualified evaluators granted sufficient access to assess risks.

This followed Anthropic CEO Dario Amodei’s suggestion that AI companies provide independent evaluators with employee-like access to verify their adherence to safety practices and report incidents of AI misbehaviour. Anthropic is one of the major firms developing AI models.

In the letter, the leaders also urged greater cooperation and transparency among governments and regional organisations regarding frontier AI.

This includes developing and coordinating common standards, shared reporting of serious safety incidents, and ensuring countries across all regions have access to scientific capacity, expertise and trusted evaluation for the technology.

The statement called on UN member states to also explore creating an international institution that would set standards for frontier AI, enable verification and convene states when capability thresholds were crossed.

In a Facebook post on Tuesday, Mr Wong said effective safeguards for AI require international cooperation given cross-border risks.

“We must continue to find practical ways to build common ground and bring more countries along,” he said, adding that the technology held tremendous promise but must remain under human control and oversight.

Singapore has launched several initiatives to build safeguards for AI, including the Singapore AI Safety Institute – which helps develop evaluation tools and common best practices for the technology.

Minister for Digital Development and Information Josephine Teo said on Tuesday that it was “encouraging” that the institute had been highlighted in an article published by OpenAI, another major AI firm, on building AI safety standards.

In a LinkedIn post about Mr Wong’s announcement, Mrs Teo said international cooperation for AI safety must also be backed by practical work.

The Singapore AI Safety Institute’s efforts to contribute to global safety standards reinforce a simple principle, she added.

“As AI capabilities advance, AI safety capabilities must advance alongside them.”

Concerns over AI have grown due to its rapidly increasing capabilities and a series of safety incidents.

In July, hundreds of OpenAI agents escaped containment and hacked into AI startup Hugging Face. The agents collaborated and left notes for each other, and tried to cover their tracks when they moved beyond the tasks set for them.

OpenAI confirmed earlier this month that autonomous software built on its models targeted another website before the Hugging Face episode.

Anthropic also disclosed in July that its AI model Claude breached the systems of three companies during testing after a configuration error gave it internet access.  

On Sep 12, Mr Amodei called on AI firms to slow down development, warning that building it too fast was "reckless".

Top AI companies' executives, including OpenAI chief Sam Altman and xAI's Elon Musk, backed Amodei's push for a slowdown. 

But critics have questioned if the recent calls for caution were to limit competition against the sector's leading firms. Others suggest these tech giants may wish to rein in capital spending ahead of planned initial public offerings, in order to burnish their books for investors.