SHARED INTEL Q&A: This is how ‘edge AI’ is forcing a rethink of trust, security and resilience
By Byron V. Acohido
A seismic shift in digital systems is underway — and most people are missing it.
IP Blacklist Check: How to Recover and Prevent Blacklisted IP Addresses
By Byron V. Acohido
A seismic shift in digital systems is underway — and most people are missing it.
While generative AI demos and LLM hype steal the spotlight, enterprise infrastructure is being quietly re-architected, not from the cloud down, but from the silicon up.
As AI use cases move from experimentation to deployment, a new layer of compute is taking shape. Behind the scenes, semiconductor companies like Infineon are already adapting embedded systems and edge architectures to handle persistent inference workloads — AI that must run in real time, in the field, at the device level.
Silicon-level trust
That transformation, however, is unfolding in parallel with more familiar challenges. Critical sectors are still struggling to patch basic vulnerabilities and navigate regulatory shifts. That tension was front and center at Infineon Technologies’ OktoberTech 2025, where Thomas Rosteck joined the stage to frame what’s really at stake.
Infineon has already begun engineering silicon-level trust features in anticipation of this shift. Last Watchdog sat down with Rosteck — division president of Connected Secure Systems — to drill down on why hardware-rooted trustis more essential than ever, and what needs to change across the device ecosystem. Here’s that discussion, edited for clarity and length:
LW: At OktoberTech, you mentioned that edge AI is bringing “cloud-level workloads to embedded devices.” What does that shift look like in practice?
Rosteck: We’re seeing a big jump in compute and memory requirements — not just in servers, but all the way out to the edge. Embedded systems are being asked to do tasks like image classification or anomaly detection that used to happen in the cloud. That means they need more performance, better energy efficiency, and critically, built-in security. It’s not about patching after deployment anymore — it’s about securing the lifecycle end to end, starting at the hardware level.
LW: What exactly does Edge AI refer to and why is that significant?
Rosteck: Edge AI means moving inference workloads away from centralized cloud servers and closer to where the data is actually generated — on the device itself. That’s a fundamental change. It allows AI models to deliver real-time insights without needing to send data to the cloud and wait for a response.
This is especially valuable in scenarios where latency, bandwidth, or data privacy are concerns. Think of a factory floor, a vehicle, or a wellness device: you want AI decisions to happen instantly and locally. That’s why we’re seeing a shift toward embedded systems and hardware that can support persistent, lightweight inference right at the edge.
That said, the edge isn’t replacing the cloud — both are still essential. But for an increasing number of real-world applications, the edge is where key parts of the decision-making can now happen.
LW: What does that mean for decision-makers triaging risk?
Rosteck: It changes the whole model. When AI inference is running on the edge device itself, you can’t rely on traditional perimeter defense. You have to trust that the device hasn’t been tampered with.
So we’re pushing for things like immutable IDs, sometimes called hardware root-of-trust, at the silicon level and hardware-based attestation. These features let systems prove they are who they say they are. That’s the foundation for secured updates, secured communication, and ultimately, trustworthy AI.
LW: You made a point about liability being a shared burden — can you unpack that?
Rosteck: Today, if a device gets compromised, people often blame the integrator or the CISO. But we think the burden needs to shift across the whole chain — from chipmakers to OEMs to service providers. That’s the only way to achieve real end-to-end trust. If any one layer is weak, the whole system is exposed.
LW: In our conversation, you spoke about the “barriers between the cloud and the edge disappearing.” What are the implications of that?
Rosteck: The line is blurring. AI models are trained in the cloud, then deployed to devices in the field. Data flows in both directions — telemetry from the edge refines the models in the cloud. But if you don’t secure the endpoints, you contaminate the whole loop. That’s why edge-to-cloud security is a continuous trust problem. It’s not just about one point in time.
LW: You made a compelling analogy: that just like firewalls became standard in enterprise IT, hardware trust anchors should become default in embedded design. Why hasn’t that happened yet?
Rosteck: It’s partly awareness, partly cost. Security is invisible until it fails. But now, with regulatory pressure and AI-driven automation, we’re seeing a shift. Devices that can’t prove their trustworthiness may not be allowed on critical networks. That’s a wake-up call.
LW: What’s one thing you wish more people understood about AI infrastructure?
Rosteck: That it’s not just about data and models. It’s about the physical hardware that runs them. If you don’t start with a secured foundation, everything you build on top is at risk. Trust must be rooted in the device — not assumed from the cloud.
Acohido
Pulitzer Prize-winning business journalist Byron V. Acohido is dedicated to fostering public awareness about how to make the Internet as private and secure as it ought to be.
(LW provides consulting services to the vendors we cover.)
December 17th, 2025 | Q & A | Top Stories
*** This is a Security Bloggers Network syndicated blog from The Last Watchdog authored by bacohido. Read the original post at: https://www.lastwatchdog.com/shared-intel-qa-this-is-how-edge-ai-is-forcing-a-rethink-of-trust-security-and-resilience/
About Author
