As enterprises speed up the deployment of LLMs and agentic workflows, they’re hitting a crucial infrastructure bottleneck: the container base photos powering these functions are riddled with inherited safety debt.
Echo, an Israeli startup, is saying a $35 million in Collection A funding at this time (bringing its to-date complete to $50 million in funding) to repair this by essentially reimagining how cloud infrastructure is constructed.
The spherical was led by N47, with participation from Notable Capital, Hyperwise Ventures, and SentinelOne. However the true story is not the capital—it is the corporate’s formidable purpose to exchange the chaotic open-source provide chain with a managed, “secure-by-design” working system.
The Hidden Working System of the Cloud
To grasp why Echo issues, you first have to know the invisible basis of the fashionable web: container base photos.
Consider a “container” like a transport field for software program. It holds the appliance code (what the builders write) and all the things that code must run (the “base picture”). For a non-technical viewers, one of the simplest ways to know a base picture is to match it to a brand-new laptop computer. While you purchase a pc, it comes with an Working System (OS) like Home windows or macOS pre-installed to deal with the fundamentals—speaking to the exhausting drive, connecting to Wi-Fi, and working packages. With out it, the pc is ineffective.
Within the cloud, the bottom picture is that Working System. Whether or not an organization like Netflix or Uber is constructing a easy internet app or a posh community of autonomous AI brokers, they depend on these pre-built layers (like Alpine, Python, or Node.js) to outline the underlying runtimes and dependencies.
Right here is the place the danger begins. In contrast to Home windows or macOS, that are maintained by tech giants, most base photos are open-source and created by communities of volunteers. As a result of they’re designed to be helpful to everybody, they’re typically full of “bloat”—a whole lot of additional instruments and settings that the majority firms do not really want.
Eylam Milner, Echo’s CTO, makes use of a stark analogy to elucidate why that is harmful: “Taking software program simply from the open supply world, it is like taking a pc discovered on the sidewalk and plugging it into your [network].”
Historically, firms attempt to repair this by downloading the picture, scanning it for bugs, and trying to “patch” the holes. However it’s a dropping battle. Echo’s analysis signifies that official Docker photos typically include over 1,000 recognized vulnerabilities (CVEs) the second they’re downloaded. For enterprise safety groups, this creates an unimaginable sport of “whac-a-mole,” inheriting infrastructure debt earlier than their engineers write a single line of code.
The “Enterprise Linux” Second for AI
For Eilon Elhadad, Echo’s co-founder and CEO, the business is repeating historical past. “Precisely what’s occurred prior to now… all people run with Linux, after which they transfer to Enterprise Linux,” Elhadad instructed VentureBeat. Simply as Pink Hat professionalized open-source Linux for the company world, Echo goals to be the “enterprise AI native OS”—a hardened, curated basis for the AI period.
“We see ourselves within the AI native period, the inspiration of all the things,” says Elhadad.
The Tech: A “Software program Compilation Manufacturing unit”
Echo isn’t a scanning device. It doesn’t search for vulnerabilities after the actual fact. As an alternative, it operates as a “software program compilation manufacturing facility” that rebuilds photos from scratch.
In response to Milner, Echo’s strategy to eliminating vulnerabilities depends on a rigorous, two-step engineering course of for each workload:
-
Compilation from Supply: Echo begins with an empty canvas. It doesn’t patch current bloated photos; it compiles binaries and libraries straight from supply code. This ensures that solely important parts are included, drastically lowering the assault floor.
-
Hardening & Provenance (SLSA Degree 3): The ensuing photos are hardened with aggressive safety configurations to make exploitation troublesome. Crucially, the construct pipeline adheres to SLSA Degree 3 requirements (Provide-chain Ranges for Software program Artifacts), making certain that each artifact is signed, examined, and verifiable.
The result’s a “drop-in alternative.” A developer merely adjustments one line of their Dockerfile to level to Echo’s registry. The appliance runs identically, however the underlying OS layer is mathematically cleaner and freed from recognized CVEs.
AI Defending In opposition to AI
The necessity for this degree of hygiene is being pushed by the “AI vs. AI” safety arms race. Dangerous actors are more and more utilizing AI to compress exploit home windows from weeks all the way down to days. Concurrently, “coding brokers”—AI instruments that autonomously write software program—have gotten the primary turbines of code, typically statistically choosing outdated or weak libraries from open supply.
To counter this, Echo has constructed a proprietary infrastructure of AI brokers that autonomously handle vulnerability analysis.
-
Steady Monitoring: Echo’s brokers monitor the 4,000+ new CVEs added to the Nationwide Vulnerability Database (NVD) month-to-month.
-
Unstructured Analysis: Past official databases, these brokers scour unstructured sources like GitHub feedback and developer boards to determine patches earlier than they’re broadly printed.
-
Self-Therapeutic: When a vulnerability is confirmed, the brokers determine affected photos, apply the repair, run compatibility checks, and generate a pull request for human evaluate.
This automation permits Echo’s engineering crew to take care of over 600 safe photos—a scale that may historically require a whole lot of safety researchers.
Why It Issues to the CISO
For technical decision-makers, Echo represents a shift from “imply time to remediation” to “zero vulnerabilities by default.”
Dan Garcia, CISO of EDB, famous in a press launch that the platform “saves at the least 235 developer hours per launch” by eliminating the necessity for engineers to analyze false positives or patch base photos manually.
Echo is already securing manufacturing workloads for main enterprises like UiPath, EDB, and Varonis. As enterprises transfer from containers to agentic workflows, the power to belief the underlying infrastructure—with out managing it—could be the defining attribute of the following era of DevSecOps.
Pricing for Echo’s resolution isn’t publicly listed, however the firm says on its web site it costs “primarily based on picture consumption, to make sure it scales with the way you truly construct and ship software program.”