Powering secure AI at the Edge: What you need to know before it’s too late

PCs with AI smarts built right into their silicon chips were transforming how we worked, empowering us to be safer and smarter no matter where we were in the world – or even if we had internet access.

But even as we adopted these new modes of work, challenges such as ‘Shadow AI’, elevated cyber risks, and user acceptance loomed large.

Cyber security leaders ranked internal AI adoption among their top risks, especially as it related to workers deploying AI on public platforms without effective safeguards or organisational management oversight. Additionally, a tendency of end-users to perceive cyber security as solely IT’s responsibility contributed to flat-footed cyber teams prematurely burning out.

As part of a regular webinar series, iTnews invited leading experts from chipmaker Intel and Datacom, one of Australia’s biggest IT solutions providers, to share insights on how business leaders could level up for the coming age of AI while defending against emergent cyber attacks.

Shadow AI influences emergent cyber threat landscape

Individual end-users blindly leaping into artificial intelligence before their employers erected appropriate guardrails was a leading cause of disquiet among Australian cyber security and business leaders, said David Stafford-Gaffney, Datacom’s associate director of cybersecurity. This was especially the case for the 16–22 per cent of ‘innovators and early adopters’ who were “sprinting towards AI”, he said.

“Users may be unknowingly uploading or exposing sensitive data to these AI platforms [or] training AI models with corporate information,” Stafford-Gaffney said.

Examples included ChatGPT integrated into PowerPoint and users linking ChatGPT to their Microsoft OneDrives.

Stafford-Gaffney said the rise of Shadow AI mirrored early cloud adoption bedevilled by misconfigured cloud storage leading to data breaches.

“We’re at that same nexus with shadow AI where we need to start thinking about how we anticipate and manage AI usage and awareness is the first part.”

He recommended organisations instil in workers “healthy paranoia” as a bulwark, especially as malicious attackers used AI for “better linguistic accuracy and regional flair” to conduct more convincing operations.

“Ask questions: ‘Am I expecting this email?’; ‘Does the sender match what I know?’,” he said.

‘No silver bullet’ but silver lining for users abandoning cloud AI

User training aside, organisations should not rest on their laurels, Stafford-Gaffney said.

“It’s not a silver bullet. We need to lift our focus on detection [because] prevention isn’t going to prevent everything; AI-powered attacks [will] get through.”

Organisations needed to invest in detection tools and cut mean time to detect and respond, he said.

Ultimately, organisations must “commit to being match fit” to repel AI-powered threats on endpoints and flaws in user behaviour, he said.

This balanced approach – combining user vigilance with robust technical safeguards – was key to preventing users from becoming unintended vulnerabilities.

But Shadow AI, when harnessed correctly, could be a powerful driver for organisational efficiency and innovation. Datacom research found that nine in 10 Australian employers encouraged their workers to use AI to save time (74%) and boost productivity (56%).

“[AI features] in Windows 11 or powered by the device … remove friction from everyday tasks,” said Tyron Hancock, Datacom Windows 11 presales specialist.

He said ‘Edge AI’, or AI embodied in silicon on the device and leveraged locally through the operating system, was a potential productivity boon for harried workers.

“We’re seeing faster turnaround times on research and summarising meeting notes, and even [real-time] translation” through Copilot+ PCs, said Hancock. He spotlighted Windows 11’s ‘Recall’ feature – a digital memory of a user’s activity – as a productivity boon when writing monthly reports.

How security-on-a-chip and ‘Edge AI’ enhance cyber defences

For IT executives and business decision-makers, Intel’s ‘security in silicon’ chip architectures – particularly within the Intel vPro platform – were a robust foundation to elevate user security and productivity, especially when leveraging Intel’s Neural Processing Unit (NPU).

“Security is fundamental to all of our platforms,” said Dino Strkljevic, Intel regional director of consumer and retail.

The NPU offloaded security tasks from the CPU to create a “very balanced system that has much longer battery life and performance as well as improved security”, he said.

Intel’s testing observed “an additional one to two hours of battery life” on their heavily used devices: “It’s ensuring significant privacy but we’re also addressing performance needs”.  

Intel had enhanced its on-chip security detection to peer into processes as they were running to determine if they were safe. What Intel called Threat Detection Technology (TDT) used AI and CPU telemetry to detect anomalies and potential threats such as ransomware, cryptojacking, and software supply chain attacks. 

“Which increased the detection rate of unusual behaviour [like] phishing attacks by an additional 20–25 per cent,” Strkljevic said.

He added that Intel engaged with some 200 independent software vendors to hunt down threats before they became wider problems in the software ecosystem, further enhancing security.

Better together – How Datacom makes a difference for customers stuck on Windows 10

And for IT executives and business decision-makers grappling with the impending Windows 10 End of Support (EOS) deadline on 14 October, Datacom had a structured, phased approach to facilitate a smooth transition to Windows 11. Datacom’s Hancock said it helped clients to rapidly adopt a compliant and supported operating system with the bonus of advanced AI capabilities.

“Datacom is the exclusive Microsoft Strategic Refresh Initiative partner here in Australia and we’ve been helping customers prepare for Windows 11 across the country for the past 12 months,” Hancock said.

At the time of the webinar recording in late June, fewer than 80 business days were left until the EoS deadline. Hancock urged organisations to act swiftly to avoid “increased risks in cyber security [and] extended support cost”.

Accelerating clients’ transition was Datacom’s ‘Paving the Way to Windows 11’, a free proof-of-concept program run over the following three phases:

Phase 1 – Discovery (readiness report): Typically held over two weeks, it leveraged Datacom’s proprietary ‘Insights’ tool to assess the client’s device fleet for Windows 11 hardware compatibility, including spotting devices ineligible for upgrade.

Phase 2 – Pilot & Testing PoC: Over four weeks, Datacom introduced a client pilot group of at least 25 users to Windows 11 while a support team ironed out emerging issues to build confidence for widespread deployment.

Phase 3 – Copilot+ PC PoC (Enabling AI at the Edge): For organisations that had switched to Windows 11, this last phase “unlocks what AI can really do at the end-point”, Hancock said. Datacom gave Copilot+ devices to individual users to experience how AI in PCs (‘at the Edge’) could improve their daily lives through features such as Recall, Live Captions and Translation, and Cocreator in Paint. Hancock: “All of this runs on the device, thanks to the NPU, which means faster performance, better privacy and no need to be constantly connected to the internet.” And it helped decision-makers understand how these advanced features operated in practice, enabling them to confidently shape their future endpoint strategy.

By leveraging this structured, data-driven approach, Datacom helped organisations navigate the complexities of the Windows 11 transition while strategically incorporating the benefits of AI at the edge, Hancock said.

Learn how Datacom can help your organisation embrace AI and transition to Windows 11 with confidence at Datacom.com