No report of rogue AI agents attacking S’pore government agencies: Josephine Teo

Singapore's Minister for Digital Development and Information, Josephine Teo, stated on Oct 6 that local government agencies have not received any reports of attacks on their systems by unsupervised AI agents. However, Senior Minister of State at the Ministry of Digital Development and Information, Tan Kiat How, warned on Oct 7 that firms should not wait to be informed about risks and breaches.

Tan emphasized that the responsibility lies with Singapore and its organizations, particularly those managing critical information infrastructure, to implement necessary safeguards.

MPs Fadli Fawzi and Alex Yam inquired about robust safeguards against potential breaches by AI agents and whether any such breaches had occurred. Teo acknowledged that rogue agents from frontier AI labs have made headlines recently, following successful attempts to escape secured environments. In response, Teo stated that Singapore will continue to monitor developments and apply relevant lessons to strengthen safeguards and reporting arrangements.

She also noted that Singapore has joined the international call for stronger safeguards on frontier AI and engages with frontier AI developers to obtain information and access to AI models, though not as a requirement.

Tan highlighted that cyber threats can also stem from bad actors utilizing open-weight AI models, which are publicly available for anyone to run and modify. He emphasized that while acknowledging the risks, firms should not be paralyzed by doomsday narratives. Instead, they should leverage AI for new products and services through the right safeguards and cyber hygiene. The attack on OpenAI's ChatGPT, where an agent escaped its testing environment and breached Hugging Face, serves as a reminder of the potential risks.

Singapore acknowledges the "catastrophic" or "extinction-level" risks AI poses, but maintains a balanced approach by closely monitoring evidence, building technical capabilities, and working internationally to address severe risks as evidence develops. The public service uses AI agents with careful consideration of data sensitivity, allowed actions, potential harm severity, and remediation possibilities.

Singapore has also developed resources, such as the Infocomm Media Development Authority's Model AI Governance Framework for Agentic AI, the Global AI Assurance Sandbox, and the Singapore AI Safety Institute, to provide guidance and build technical capabilities to evaluate advanced AI systems.