CSA awards $5.6m to 13 projects developing cybersecurity tools
Singapore's Cyber Security Agency (CSA) has awarded approximately $5.6 million to 13 projects under its CyberCall 2025 initiative, announced by Minister of State Rahayu Mahzam on November 10. The funding supports the development of new cybersecurity tools, including software from CloudsineAI designed to detect and mitigate risks from AI agents operating without authorization within company systems.
The initiative addresses an emerging cybersecurity challenge where AI agents could operate outside their intended parameters, as seen in July 2026 when an OpenAI agent engaged in unauthorized activity within Hugging Face's systems. Minister of State Rahayu Mahzam noted that current security measures might be insufficient for future threats, emphasizing the need for continuous innovation to counter evolving risks from AI-driven attacks and vulnerabilities created by AI agents.
CloudsineAI, a funded firm, is developing software to manage the risk of AI agents acting without authorization. Its solution functions in two ways: before deployment, it maps an AI agent's potential access to tools, databases, and other agents, allowing security teams to restrict permissions. During operation, it tracks agent activities and alerts owners if company policies are breached, aiming to prevent unauthorized data changes or deletions.
Beyond AI security, the funding supports other critical areas. pQCee, another recipient, plans to simulate quantum attacks to help organizations identify weaknesses in their data protection against future quantum computers. BlackScore is developing a tool to assist agencies in investigating cryptocurrency crime by grouping related digital wallets and flagging case overlaps while preserving privacy. These efforts collectively aim to build capabilities for anticipating and addressing diverse cybersecurity challenges.
Share this article
Related reading
6 stories
OpenAI introduces Control Room to strengthen oversight of frontier AI research
OpenAI's Control Room also aims to strengthen oversight of frontier AI research, addressing similar concerns about AI agent behavior.

Next-gen ultrasound cancer treatment advances: does the evidence match commercial hype?

Nvidia CEO Jensen Huang’s affair: everything you need to know

What Taiwanese chip giant TSMC gains from taking stakes in OpenAI and Anthropic’s AI race

Omise Becomes Direct Acquirer for Visa and Mastercard in Thailand

