Atsign launches EdgeAI with identity-based kill switch
Atsign's EdgeAI package gives distributed AI fleets an identity-based kill switch and hardware-backed encryption, with throughput up to 6 Gbps on Intel Core Ultra chips.

Why it matters
- Atsign launched Atsign EdgeAI on 17 November 2026 for organisations running distributed AI and connected systems.
- EdgeAI bundles device, gateway and edge software with the Atsign Control Centre, which sets communication rules by identity rather than IP address.
- An identity-based kill switch lets administrators cut off compromised devices, applications or AI agents without firewall changes.
- In Atsign's testing, encryption reached up to 6 Gbps on Intel Core Ultra versus a ~100 Mbps software baseline, and up to 5 Gbps on Intel Xeon versus ~57 Mbps.
- Atsign and Intel have published a smart transportation brief on hardware-backed security features; Intel's Renu Navale cited scaling across cities and critical infrastructure.
Atsign launched Atsign EdgeAI on 17 November 2026, a package aimed at organisations running AI systems and connected devices across separate networks. It bundles device-side software with a management layer called the Atsign Control Centre, adds an identity-based kill switch for compromised devices and AI agents, and — in Atsign's own testing — pushed encryption throughput to up to 6 Gbps on Intel Core Ultra processors, roughly 60 times a software-only baseline.
The product lands on a real operational fault line. Companies pushing AI out of centralised cloud environments now run fleets of machines across factories, vehicles, depots and cities, and each site often carries its own mix of private, public, cellular and customer-managed networking. The result is that a pilot that works on one site can take months of security and networking rework to roll out at the next one.
What problem is Atsign EdgeAI meant to solve?
Atsign built EdgeAI on its existing platform. The company describes it as a way to connect devices, applications and AI systems across different networks while reducing deployment complexity, and it is pitching the package into edge AI, physical AI, industrial and operational technology, smart transport, robotics, critical infrastructure and remote operations.
The core difficulty is fragmentation. A multi-site AI deployment touches private LANs, public links, carrier networks and infrastructure the customer itself manages. Each new site becomes a networking project before it becomes an AI project, and security teams have to expose internal systems or open holes in firewalls to let anything connect.
"Getting a pilot up and running is one thing. Scaling it across dozens or hundreds of sites, each with different networks and security requirements, is where things get difficult," Aparna Rayasam, Chief Executive Officer of Atsign, said. "We built Atsign EdgeAI to make those deployments simpler and more secure, without forcing companies to expose internal systems or rework their networks at every location."
That quote captures the commercial pitch in one line: the product is sold against scaling pain, not against a single technical feature.
How does identity-based access work?
EdgeAI combines software running on devices, gateways, applications and edge systems with the Atsign Control Centre, the management layer administrators use to see across distributed environments.
The control centre does not base access rules on where something sits on a network. It ties communication rules to identity rather than network location or IP addresses.
The mechanism works like this:
- Administrators assign cryptographic identities to users, devices, applications, services, processes and AI agents.
- Those identities — not IP addresses or VLAN membership — determine which participants may talk to which specific resources.
- The same rules apply at every site, so adding a location does not require a fresh network design.
Atsign argues that this removes site-by-site network redesign when connecting distributed systems. The control centre also gives administrators visibility over the whole estate, which matters when a fleet spans hundreds of sites that no one person can inspect directly.
What does the kill switch actually do?
The feature Atsign is highlighting hardest is an identity-based kill switch. When administrators detect a compromised device, application or AI agent, they can cut off or restrict its communications without waiting for firewall changes or a wider network reconfiguration.
That distinction matters for agentic deployments. Software agents act on their own, across machines and networks, and a traditional containment response — change the firewall, pull a VLAN, wait for a maintenance window — is too slow for a system that may already be making decisions. Revoking an identity takes effect at the point of communication instead of at the edge of the network.
Takeaway: the product treats identity as the control point, so containment becomes a lookup rather than a reconfiguration project.
What does the Intel work add?
Atsign also pointed to work with Intel on securing communications between autonomous systems in edge environments. The two companies have published a smart transportation brief covering the use of Intel hardware-backed security features alongside Atsign's software.
According to Atsign, the Intel-based approach includes three specific elements:
- Encryption acceleration
- Hardware-backed key generation
- Protected key storage
The performance numbers came from Atsign's testing. Encryption throughput reached up to 6 Gbps on Intel Core Ultra processors, against a software-only baseline of about 100 Mbps. On Intel Xeon processors, throughput reached up to 5 Gbps, against a baseline of about 57 Mbps. Atsign put those figures at increases of up to 60 times on Intel Core Ultra chips and up to 88 times on Intel Xeon processors.
Atsign says the results show that hardware-assisted cryptography can make secure communications more practical in distributed edge settings using agentic AI systems. The logic is straightforward: if encryption runs in hardware, the cost of encrypting every message between autonomous systems drops, and security stops being the reason a rollout slows down.
"Edge AI deployments only deliver value if organizations can trust and scale them," Renu Navale, VP and GM, Health Cities Division, Edge Computing Group at Intel, said. "With Intel hardware-backed security and encryption acceleration, customers can protect communications between autonomous systems without slowing them down. That means stronger security, faster rollouts, and the confidence to scale Edge AI across cities, transportation networks, and critical infrastructure."
Where does this sit in the wider security market?
The announcement reflects a sector-wide push to secure AI systems operating outside conventional data centre environments. As more organisations deploy software agents, connected machinery and remote monitoring systems, vendors are trying to address the risks that come when those systems sit exposed to the public internet or depend on fragmented networking arrangements.
Atsign's answer is architectural. The company argues its approach reduces exposure by keeping protected systems out of direct public reach and encrypting communications end to end between authorised participants. The underlying platform can be used either as a packaged product like EdgeAI or embedded into applications, devices and AI agents through software development tools.
The emphasis on identity-based controls over location-based network access also reflects a broader shift in cybersecurity towards zero-trust models. In distributed AI environments, access decisions attach to verified participants and defined relationships rather than assuming a system is trusted because it sits inside a given network boundary. That shift is not Atsign's invention, but it names the trend the product is built for.
What to watch
Two questions sit behind the launch. The first is whether operators will accept identity-based control over their existing network management, since EdgeAI asks them to give up location-based rules they have maintained for years. The second is whether the Intel throughput advantage holds up outside Atsign's own tests, and whether hardware-assisted cryptography becomes a default requirement for agentic systems in the field.
For now, the concrete claim is a modest one: the platform lets an administrator revoke an agent's access instantly across sites, and the encryption numbers suggest the security tax on those communications no longer has to be paid in throughput.
Source: itbrief.com.au
More from Sophie Lindqvist
Show full bio
Staff writer covering regulation, policy and their impact on artificial intelligence and machine learning at AI In Context.
749 articles
Related articles
- Coalition plugs OpenAI, Anthropic, Gemini and Cursor into Wirespeed
- Arista ships four encrypted routing switches for distributed AI inferencing
- Intel's Edge AI Push Cannot Replace Its Shrinking PC Business
- Axway launches Amplify Platform to control AI costs and data risk
- Cloudflare Brings OpenAI's GPT-5.4 to Agent Cloud for Enterprises