Flexxon awarded funding for the third time under CyberCall to tackle AI agent security at the hardware level

X-PHY AI Armor will watch AI agents at the protocol level and enforce controls at the hardware level, closing the gap left open by software-only controls.

SINGAPORE, Nov. 10, 2026 /PRNewswire/ -- Flexxon has been named an awardee of the Cybersecurity Industry Call for Innovation (CyberCall) 2025. Since 2024, CyberCall has been organised by the CyberSG Talent, Innovation and Growth Collaboration Centre (TIG Centre), a joint initiative by the National University of Singapore (NUS) and Cyber Security Agency of Singapore (CSA). The award, made under the "Cybersecurity for AI" challenge, was announced at the CyberSG Innovation Day on 10 November 2026.

The project, X-PHY AI Armor, will be a hardware-rooted secure Model Context Protocol (MCP) gateway, building on the company's hardware cybersecurity expertise. It will sit between the AI agent and the systems it reaches, inspecting every request the agent makes and looking for injected instructions and hijacked tools. Anything high-risk will be handed to the hardware, which will lock it down.

Because authorisation will be anchored in X-PHY hardware rather than in software, a stolen credential on its own will not be enough to get through. Generic API gateways inspect traffic at the network layer and can be reconfigured by anyone who has taken control of the host. X-PHY's enforcement will sit below that: authorisation for high-risk operations will be verified by the device, and the audit record will be written to the hardware. An attacker who controls the host will not be able to alter the decision or erase the trace of it.

Ms Camellia Chan, Co-founder and CEO of Flexxon, said, "Agents are useful precisely because they act without waiting for a person, and that is exactly what makes them dangerous. An agent following a malicious instruction is indistinguishable from an agent doing its job. You cannot solve that by asking software to watch itself more carefully - the judgement has to sit somewhere an attacker cannot reach. Our thanks to CSA and to the TIG Centre, and to the judging panel, for backing us once again and for the collaboration that makes much-needed solutions like ours possible."

A third award, and a new frontier

The award is Flexxon's third under the same initiative. The company was one of eight selected from 72 submissions in CyberCall 2019 for its X-PHY Solid State Drive (SSD), and an awardee of the 2021 edition for its server security module.

The previous projects secured data at rest and the systems around it, while AI Armor will address a problem that did not exist when they were funded. In the last two years, agentic AI has moved from demonstration to deployment, and the security model has not kept pace with it.

Agents hold credentials, reach databases and APIs, and act at machine speed, and the controls meant to supervise them run on the very software layers an attacker is trying to take over.

The scale of the AI agent problem

An AI agent works by holding credentials and using them - reading files, querying databases, calling tools and passing instructions to other agents, without a person checking each step. Those instructions arrive as text, and text can be poisoned. A single injected line in a document, a web page or another agent's reply can redirect the agent, while every action it then takes remains technically authorised.

CSA's Singapore Cyber Landscape 2025/2026, published in June this year, warned that agentic AI systems "could potentially automate significant parts of the cyber kill chains by chaining exploit steps together autonomously at speed", and that "intrusions that previously unfolded over days can now proceed within hours, narrowing the gap between criminal and state-level operators". The report cites OpenClaw, "the open-source agentic AI technology that was abused to breach development pipelines at scale", as a case in point.

Singapore organisations are meeting that on uneven ground. CSA's Singapore Cybersecurity Health Report 2025, published in October 2026, surveyed 2,005 businesses and non-profits and found almost nine in 10 had encountered at least one cybersecurity incident during the year. Organisations are adding AI agents and their associated threats to that environment.

In fact, Gartner forecasts that up to 40 per cent of enterprise applications will ship with task-specific AI agents integrated by the end of 2026, up from fewer than 5 per cent a year earlier.

The organisations that will feel this most immediately are those that have given agents access to something worth taking: the financial services, healthcare and government sectors, and the data centre operators running the infrastructure beneath them.

"Securing AI agents will be a major cybersecurity challenge in the coming years, and it is exactly the kind of problem CyberCall was established to address. We are pleased to support Flexxon as it builds on its expertise to develop X-PHY AI Armor to tackle this emerging challenge," said Mr Willis Lim, Executive Director, TIG Centre.

Building the hardware-anchored ecosystem

Flexxon will develop X-PHY AI Armor with support from CyberCall, building on its existing expertise in hardware-level cybersecurity. The solution aims to help organisations harness the benefits of AI agents while safeguarding the security and integrity of their data and systems.

SOURCE Flexxon