SandGrain and Axelera AI join forces to advance trusted AI at the edge

As artificial intelligence moves increasingly to the edge, trust becomes fundamental. It is no longer enough to know that an AI model can process data quickly and efficiently. Organisations also need confidence in where that data comes from, which devices can be trusted, and that the AI models themselves remain protected.

SandGrain and Axelera AI are joining forces to advance trusted and secure AI at the edge, combining Axelera AI’s high-performance edge AI technology with SandGrain’s hardware-anchored trust.

Together, the companies will explore joint opportunities, technical integration and go-to-market activities around trusted Edge AI, making it easier for customers to deploy AI they can trust—with verified data sources and protected models running where their data is created.

Bringing trust to Edge AI

AI systems operating at the edge increasingly make decisions close to where data is generated — whether that data comes from cameras, sensors, machines or other connected devices.

But the value of those decisions ultimately depends on whether the underlying data and devices can be trusted.

SandGrain’s CyberRock technology establishes hardware-anchored device identity, enabling devices and data sources to authenticate themselves before access is granted. In an Edge AI environment, this creates the foundation for establishing trust in the source of the data being processed.

Combined with Axelera AI’s high-performance AI processing technology, this trust mechanism can also be used to protect the AI model itself. In a joint implementation, a camera is authenticated and bound to the SandGrain Trust Anchor IC. Only after successful authentication is the decryption key for the protected AI model released, helping ensure that an extracted model cannot simply be reused elsewhere.

As a first tangible outcome of the collaboration, SandGrain and Axelera AI are developing a joint reference implementation that brings these capabilities together in a live multi-vision Edge AI environment.

Live multi-vision Edge AI demonstration

SandGrain and Axelera AI will present a live multi-vision Edge AI demonstration combining SandGrain’s hardware-anchored device authentication with Axelera AI’s AI processing technology.

The demonstration will show multiple computer-vision streams being processed at the edge while establishing trust in the data source and protecting the AI models used to process that data.

The result is an architecture in which AI processing does not simply rely on receiving data — it can establish whether the source of that data can be trusted before the protected AI model is made available.

This becomes increasingly important as AI moves from centralised environments towards distributed edge architectures, where systems may need to make autonomous or time-critical decisions without continuous reliance on a central cloud infrastructure.

Building trusted AI at the edge

The collaboration between SandGrain and Axelera AI creates a foundation for broader technical and market cooperation around trusted Edge AI. Alongside the joint reference implementation, the companies will explore further customer opportunities and ways to bring the combined proposition to demanding real-world environments.

By combining high-performance Edge AI with hardware-anchored trust, SandGrain and Axelera AI aim to address a fundamental question for the next generation of distributed AI systems:

If AI is increasingly making decisions at the edge, how do we know that the devices, data and models behind those decisions can be trusted?

The collaboration brings together two complementary European technologies to help answer that question — establishing trust at the point where data is generated, processed and turned into decisions.