NewCore Raises $66M to Build Identity Infrastructure for the AI Agent Workforce
Table of Contents
The Shift From Tools to Digital Employees
The traditional boundary between software and staff is blurring. For decades, enterprises have viewed AI as a tool—a calculator or a spreadsheet that requires a human to operate it. But the emergence of autonomous agents, capable of executing multi-step workflows and making independent decisions, has shifted the paradigm. When a coding agent like Devin or an automated analyst from McKinsey operates across a network, it is no longer behaving like a piece of software; it is behaving like an employee.
This shift creates a massive security vacuum. Most companies currently manage non-human access through service accounts or machine credentials—static keys that are often over-privileged and rarely rotated. NewCore, a cybersecurity startup that recently emerged from stealth with $66 million in seed funding, is betting that this legacy approach will collapse as AI agent deployment scales.
- The Funding: Led by Cyberstarts, with Index Ventures and Evolution Equity Partners.
- The Valuation: $300 million post-investment.
- The Goal: Replacing legacy identity platforms with a system designed specifically for a hybrid workforce of humans and AI.
The urgency is underscored by the scale of adoption. McKinsey has already integrated 25,000 AI agents into its operational fabric, while Goldman Sachs has experimented with AI agents in coding roles. As these entities move from ‘experimental’ to ‘essential,’ the question of who (or what) is accessing sensitive data becomes a critical failure point for the modern CISO.
The Failure of Legacy Identity Platforms
Current Identity and Access Management (IAM) giants like Okta and Microsoft Entra were built for a world where a human enters a password, perhaps verifies it with a thumbprint, and is granted access to a set of apps. AI agents don’t have thumbs, and they don’t experience ‘session timeouts’ the way humans do. When an AI agent is granted a broad service account credential, it often possesses ‘god-mode’ permissions that far exceed what a human employee would have, creating a significant attack surface for hackers.
Zohar Alon, NewCore’s co-founder and CEO, argues that legacy systems are essentially being patched to handle AI, rather than being rebuilt for it. Alon, who previously scaled Dome9 before its acquisition by Check Point, suggests that adding ‘agentic’ features to a 20-year-old identity framework is akin to putting a jet engine on a horse carriage. The underlying architecture cannot support the volatility, scale, and autonomy of thousands of agents acting simultaneously.
The Danger of Over-Privileged Service Accounts
In a standard enterprise setup, a service account is used for automated tasks. Because rotating these keys often breaks the automation, they are frequently left unchanged for years. If a malicious actor compromises a service account used by an AI agent, they gain a permanent, invisible doorway into the heart of the network. NewCore’s approach treats AI agents as first-class identities. This means an agent has its own lifecycle: it is hired (provisioned), given specific limited permissions (scoped), and fired (revoked) just like a human staff member.
Technical Deep Dive: Split-Key Architecture and Agentic Skills
To move beyond the vulnerabilities of traditional IAM, NewCore has introduced a split-key architecture. In traditional systems, the identity provider holds the master key. If the provider is breached, every single account under its umbrella is compromised. NewCore divides the critical credentials between the customer and the platform. This means neither the provider nor the client possesses the full key in isolation, eliminating the ‘single point of failure’ that has plagued previous major cybersecurity breaches.
Integrating with the AI Ecosystem
NewCore isn’t just providing a dashboard; it’s building an integration layer for the most popular agentic tools. The company has released an “Agentic Skill” package designed for high-autonomy tools including:
- Claude Code (Anthropic): Allowing the agent to navigate repositories without requiring the developer to hardcode API keys into the environment.
- OpenAI Codex: Managing the identity of code-generation agents as they interact with production databases.
- Cursor: Enabling a managed identity for the AI-powered IDE to access internal enterprise documentation securely.
By moving credentials out of the code and into a managed identity layer, NewCore reduces the risk of ‘secret leakage’—one of the most common ways cloud environments are compromised today.
The Human-in-the-Loop Oversight Layer
One of the most contentious points in AI deployment is the loss of control. As agents become more autonomous, the ‘black box’ problem grows. NewCore addresses this by introducing a human oversight layer via a mobile application. Instead of relying on a complex admin console, employees can receive notifications on their phones to grant, review, or instantly revoke access for an AI agent in real-time.
This transforms the security model from static permissioning (where you decide once what an agent can do) to dynamic authorization (where you approve specific high-risk actions as they happen). For example, if an AI agent decides it needs to access a sensitive financial folder to complete a report, the human manager can approve that specific request for a limited window of time.
The Market Opportunity: The Rise of the ‘Digital Workforce’
The bet NewCore is making is not just on security, but on the sheer volume of AI agents. N. Chandrasekaran, Chairman of TCS, has noted that AI agents could eventually rival the size of traditional IT service workforces. If an organization with 10,000 humans suddenly adds 50,000 AI agents, the overhead of managing those identities manually becomes impossible.
Competitive Landscape: NewCore vs. The Incumbents
While Microsoft and Okta are not standing still, their primary business model relies on the stability of the human-centric identity. Pivoting to a machine-first identity requires a fundamental change in how they handle authentication tokens and session management. NewCore’s advantage is its ‘greenfield’ status—it doesn’t have to maintain backward compatibility with the legacy protocols of the 2000s.
| Feature | Legacy IAM (Okta/Entra) | NewCore Approach |
|---|---|---|
| Identity Type | Human-centric (with service accounts) | Hybrid (Humans + AI as first-class) |
| Credential Model | Centralized Master Keys | Split-Key Architecture |
| Permissioning | Static/Role-Based (RBAC) | Dynamic/Context-Based |
| Management | Admin Console | Mobile-first Human Oversight |
| AI Integration | Plugin-based/Add-on | Native Agentic Skill sets |
What This Means for the Modern Enterprise
For most businesses, the immediate implication is a shift in how they budget for AI. It is no longer enough to pay for a LLM license (like ChatGPT Enterprise or Claude for Business). Companies must now account for the governance layer. Without a system like NewCore, the deployment of AI agents is effectively a trade-off: you gain productivity, but you increase your cyber-risk profile exponentially.
For developers, this means a move away from .env files and hardcoded secrets. The goal is a world where the AI agent ‘identifies’ itself to the system, the system checks with NewCore to see if that agent is currently authorized, and access is granted on a just-in-time basis.
Addressing the Challenges of AI Governance
How does this affect data privacy?
By isolating AI identities, companies can implement stricter data silos. Instead of an agent having access to ‘everything the developer has access to,’ the agent is restricted to only the specific datasets required for the task. This limits the blast radius of a potential prompt-injection attack.
Is this compatible with existing regulations?
With the EU AI Act and other emerging frameworks, the requirement for ‘traceability’ and ‘human oversight’ is becoming law. NewCore’s audit logs—which record exactly which agent accessed what data and which human approved it—provide a ready-made compliance trail for regulators.
The Risk of Over-Reliance
While NewCore solves the identity problem, it does not solve the ‘hallucination’ problem. An authenticated AI agent can still make a catastrophic mistake. The identity layer ensures the agent has the right to do something, but it doesn’t guarantee the agent is doing the right thing. Enterprise leaders must distinguish between access control and quality control.
Frequently Asked Questions
What exactly is an AI agent identity?
An AI agent identity is a unique digital credential assigned to an autonomous AI system. Unlike a standard API key, it includes a full lifecycle, specific permissions, and a link to a human overseer, allowing the company to treat the AI as a managed entity rather than a generic piece of software.
How does the split-key architecture improve security?
In a split-key system, the secret required to authenticate an identity is divided into two parts. One part is held by the customer and one by the service provider. An attacker would need to breach both independent environments simultaneously to steal the credential, making a total system compromise significantly harder.
Can NewCore replace Okta or Microsoft Entra?
NewCore is designed to complement or evolve beyond these systems specifically for agentic workflows. While it manages identities, it focuses on the unique needs of AI agents (high frequency, autonomous action, dynamic permissions) that legacy systems struggle to handle.
Does this stop AI agents from hallucinating?
No. NewCore manages who can access the data, not what the AI does with that data. It is a security and governance tool, not a tool for improving the accuracy of the AI’s output.
Who is this technology for?
Primarily for mid-to-large scale enterprises that are moving beyond simple chatbots and deploying autonomous agents (like coding assistants or automated researchers) across their internal networks.
