Two technology companies say they have cracked a problem that has quietly limited autonomous AI systems at the network edge: encryption too slow to keep up. Atsign and Intel report that, by combining software identity controls with Intel’s built-in hardware security, they achieved an approximately 88-fold — roughly 8,700 percent — performance improvement for end-to-end encrypted communications between AI agents.
The companies detailed the work in a jointly published solution brief focused on smart transportation, where intersections, traffic corridors and mobility networks increasingly rely on distributed intelligent systems making decisions in real time. In that environment, every device, agent, data exchange and cryptographic key has to be trustworthy, or the entire system’s decisions become suspect.
Under the arrangement described by the companies, Atsign provides trusted identity management, namespace isolation, scoped data sharing, end-to-end encryption and auditability, while Intel contributes hardware-backed key generation, accelerated encryption and hashing, protected key storage, secure boot and related platform protections. The companies say this combination eliminates a major barrier to autonomous edge agents in transportation systems.
The claims come from the vendors themselves and await wider independent benchmarking. But the direction of travel is unmistakable: as AI agents move out of data centers and into cars, intersections and industrial equipment, the security plumbing — identity, encryption and hardware roots of trust — is becoming as commercially important as the models themselves.
Why Encryption Becomes a Bottleneck at the Edge
Edge computing places processing close to the place where data is created, rather than sending everything to a distant data center for a decision. In transportation, that might mean a vehicle or roadside system reacting in fractions of a second. Security cannot simply be added afterward, because messages between devices need to be verified and protected throughout their journey. Encryption and identity checks consume computing cycles and add latency. On powerful servers that overhead may be negligible, but distributed systems can involve many small devices exchanging frequent messages. If every exchange must be authenticated and encrypted, a slow security layer can undermine the very responsiveness that justified moving the intelligence to the edge. That is the operational problem the companies say their combination is designed to address.
How Software Identity and Hardware Roots of Trust Fit Together
Modern secure systems usually combine several layers. Software assigns identities to devices and applications, controls which systems may communicate, and records who accessed what. Hardware contributes a protected place to generate and store cryptographic keys, functions that accelerate common encryption operations, and checks that the device booted with trusted software. The idea is that a key locked into hardware is harder to extract or clone than a credential stored only in ordinary memory. When these layers are designed together, a system can verify both the machine and the software identity of an agent before it acts on a message. In a network of autonomous agents, that matters because a technically valid command from an untrusted source is still a threat.
What Independent Benchmarking Would Need to Show
Vendor performance claims are most useful when they can be compared on transparent terms. Independent testing would normally clarify the devices used, the workload measured, the type and length of encryption, network conditions and the baseline chosen for comparison. A large percentage improvement can mean different things depending on whether the starting point was constrained by hardware, software configuration or an inefficient reference design. For buyers in transportation and industrial settings, the deeper questions include reliability over time, ease of key management, compatibility with existing equipment and behavior when a device loses connectivity. The announcement signals where competition is moving: toward systems in which identity, encryption and hardware trust are sold as an integrated foundation for autonomous operations, rather than as separate components assembled by each customer.
Why Transportation Is the Test Case
Smart transportation systems combine vehicles, roadside sensors, traffic signals and control centers that were not always designed to operate as one secure network. Devices may remain in service for many years, communicate over changing wireless links and need updates without being taken offline for long. Those conditions make identity management more than an information technology detail. A system must know which sensor supplied a reading, which agent issued an instruction and whether either has been replaced or compromised. Faster encrypted communication would allow more of those checks to happen continuously rather than occasionally. Whether the reported improvement translates into field deployments will depend on cost, interoperability and evidence that the technology maintains its advantage outside the vendors chosen test environment.
Related reading: CyberSentriq Launches SentriqAI to Bring Enterprise-Grade Defense to Small Firms · Microsoft Puts AI Agents in Execution Containers and Ships a $2,599 RTX Spark Laptop · Starlink Mobile's Last Puzzle Piece: What Low-Band Spectrum Means for Your Phone
