Atsign says a combination of its products has now been independently tested and validated as a possible answer to a problem many companies are quietly struggling with: how to deploy AI agents without opening new security gaps in the process. The company announced that its Atsign AI Architect and Atsign Platform pairing received a Gold Award following evaluation by Broadband-Testing, one of Europe’s established independent testing facilities. The review focused on a question becoming harder for enterprise security leaders to ignore: can AI systems operate safely without exposing infrastructure in the ways traditional software often does?
That question matters more than it did even a year ago. Enterprises are experimenting with AI agents almost everywhere – customer support, software engineering, internal automation, analytics, healthcare workflows, and finance operations. The ambition is clear. So is the hesitation. Many teams discover quickly that the same networking assumptions underpinning conventional software create uncomfortable trade-offs once autonomous systems start making decisions and interacting with sensitive systems on their own.
The same tension is surfacing beyond cybersecurity, too. Warehouses and logistics environments, for example, are beginning to wrestle with a similar reality as AI moves from experimentation into everyday decision-making. Companies modernizing supply chain operations are increasingly studying how intelligent systems improve throughput, forecasting, and execution without adding operational complexity. Resources such as this AI in warehouse operations eBook are becoming useful reference points for teams trying to work out where automation creates measurable gains and where investment priorities should sit.
Atsign’s argument is fairly direct: the networking model itself needs to change.
Instead of relying on exposed inbound TCP ports and intermediary trust models, the company’s setup approaches communication differently. Atsign AI Architect provides a visual environment for designing agent behavior, while the underlying Atsign Platform handles encrypted communication through a zero-knowledge approach that avoids leaving traditional network doors open in the first place.
For developers, the appeal is practical rather than theoretical. Less infrastructure exposure to secure. Fewer assumptions around trusted connections. Fewer moving parts that become someone else’s incident later.
AI Agents Are Forcing Security Teams to Rethink Old Assumptions
The conversation around AI security has shifted noticeably in recent months. Early discussions focused mostly on data privacy, hallucinations, and governance. Now the attention is drifting toward something more foundational: what happens once agents are trusted to actually do things.
Autonomous Systems Behave More Like Operators Than Software
Traditional applications generally wait for instructions. AI agents increasingly do not. They interact with APIs, pull information from systems, trigger actions, make recommendations, and in some cases move through chains of tasks with very little human involvement.
That creates efficiency. It also introduces a level of access complexity that most security environments were not really designed for.
A broadly permissioned agent communicating through exposed infrastructure changes the risk equation quickly. A mistake made at machine speed looks very different from one made by an employee clicking the wrong button.
Aparna Rayasam, CEO of Atsign, said many organizations are still relying on security assumptions that made sense for older systems but feel increasingly fragile when applied to agentic AI.
“Standard protocols require open network ports and trusted intermediaries, inviting attackers directly into the heart of the enterprise,” Rayasam said.
Her argument is less about adding another security layer and more about removing unnecessary exposure entirely.
Real Deployments Mattered More Than Theory in Testing
One reason the validation may resonate with technical teams is that Broadband-Testing looked at functioning use cases rather than conceptual architecture diagrams.
Healthcare and Industrial Environments Were Included
The assessment reviewed two working implementations: a telemetry monitoring application connected to a KRYZ-LPFM radio transmitter and Pembrook, a secure personal AI agent.
According to the report, both environments avoided several attack paths security teams routinely worry about, including credential theft, port scanning, and man-in-the-middle attacks.
The testing group did not downplay its conclusion.
“For businesses looking to maximise their ‘ideation to adoption’ process with AI, what Atsign is offering, in terms of zero attack surface and policy control, is an absolute game changer,” the report said.
That wording will likely stand out to CISOs who continue facing pressure to accelerate AI programs while still being expected to explain exactly how those systems are protected.
Security Friction Often Slows AI More Than Technology Does
In practice, many enterprise AI projects stall for reasons that have little to do with model performance.
Teams Often Get Stuck on Infrastructure Questions
Rick Deacon, CISO and Head of Platform at NeuroVitals, said that was part of the appeal during his team’s experience building a healthcare application involving highly sensitive data.
Instead of spending days assembling cloud infrastructure and security architecture before meaningful development could begin, much of the complexity was already handled.
“Atsign removes a lot of the infrastructure setup, the things I would normally need a sophisticated architect to create on AWS or GCP just to make sure it’s secure,” Deacon said.
He added that a working MVP came together in roughly three hours – significantly faster than earlier attempts using more conventional approaches.
“The levels and layers of communication are entirely encrypted, and I have total peace of mind knowing that no one is going to see that data except the intended users,” he said.
For heavily regulated sectors, speed alone is rarely enough. The bigger question is whether teams can move quickly without creating security debt that they will spend months cleaning up later.
AI Security Is Quietly Moving Upstream
The larger takeaway from Atsign’s announcement has less to do with a single product launch and more to do with how enterprise thinking around AI security is changing.
For years, many companies treated protection as something added near the end of development – harden the system later, lock things down before release.
Agentic AI is making that model harder to defend.
When systems are capable of acting autonomously, interacting with sensitive infrastructure, and making decisions across environments, trust becomes something that has to be designed into the foundation rather than layered on after the fact.
Research and Intelligence Sources: Atsign
To participate in our interviews, please write to our CyberTech Media Room at info@intentamplify.com
