CSA awards $5.6m to 13 projects developing cybersecurity tools

SINGAPORE – AI agents given access to a company’s database could go rogue – changing or deleting records without authorisation – and the company could find out about it only when the damage has been done.

Singapore cybersecurity firm CloudsineAI is developing software that helps companies spot such risks before they deploy such artificial intelligence capabilities.

The software maps what each AI agent can reach, including tools, databases and other agents, so security teams can cut back access where needed.

It also tracks what agents do once they are running and alerts their owners if they break company policies.

CloudsineAI’s project is one of 13 that will share about $5.6 million in funding under CyberCall 2025, the Cyber Security Agency of Singapore’s (CSA) annual challenge to help companies develop and test new cybersecurity tools.

Minister of State for Digital Development and Information Rahayu Mahzam announced the award recipients on Nov 10 at CyberSG Innovation Day 2026 at Suntec Convention Centre.

She said Singapore has to prepare for new risks as it adopts new technology.

Perpetrators are already using AI to launch more sophisticated attacks at greater speed and scale, she noted, while AI agents are creating security challenges that are only beginning to be understood.

“The solutions that protect us today may not be enough for the challenges of tomorrow. We therefore need to continually innovate by developing new approaches and capabilities to keep pace with emerging threats,” she said.

The need for organisations to mitigate the risks of using AI became clear earlier in 2026.

In July, an AI agent that OpenAI was testing for its hacking abilities broke out of a restricted test environment and spent more than two days inside the systems of Hugging Face, a platform that hosts AI models and datasets. Other instances of AI going rogue were subsequently reported.

CloudsineAI is engaging companies in regulated industries like banking and finance and Government agencies to run proof of concept trials.

AI agents could outnumber human workers within a few years, said its founder and chief executive Matthias Chin.

Apart from AI security, post-quantum cryptography is another field where advances in technology could outpace existing defences.

Post-quantum cryptography uses new mathematical methods to protect data and verify digital signatures against future quantum computers that could crack some of the methods in use today.

Singapore firm pQCee, one of the other recipients of the CSA fund, plans to simulate quantum attacks to help organisations find weaknesses in how they protect their data and decide what to upgrade first.

Another recipient, BlackScore, is building a tool to help agencies investigate cryptocurrency crime. It will group related digital wallets and alert investigators when their cases overlap, while keeping sensitive case details private.

Rahayu said: “Our aim is not simply to develop more technologies, but to build the capabilities that will help Singapore anticipate and address emerging cybersecurity challenges.

“Not every idea will succeed. But by learning quickly and building on what works, we can develop the capabilities needed to stay ahead of evolving threats.”

She said Singapore also wants to be a launchpad for cybersecurity firms looking to compete overseas. Working with local customers would give firms a chance to prove their products can protect critical systems and build a track record they can take abroad.

CSA also launched CyberCall 2026, which asks companies to tackle three practical problems. It is seeking tools that can automatically detect and respond to fast-moving cyberintrusions, spot fake aircraft tracking signals without disrupting aviation operations, and help small businesses deal with cybersecurity problems on their own.

A separate open category welcomes proposals on AI security, using AI for cyber defence, preparing for quantum threats, securing cloud services and connected equipment, and protecting data privacy. Proposals are due on Jan 9, 2027.