Every business deploying AI agents this year is quietly facing a question that the mainstream security industry has not fully answered: who — or what — is that agent, exactly?

Today, Tel Aviv and San Francisco-based startup NewCore emerged from stealth with $66 million in seed funding to answer that question. Backed by Cyberstarts, Index Ventures, and Evolution Equity Partners, the company is valued at $300 million and is targeting enterprise customers who are deploying AI agents at scale and discovering that their existing identity security infrastructure was never built to handle them.

The Problem No One Planned For

For decades, identity security meant one thing: humans logging into applications. Platforms like Okta and Microsoft Entra were designed around this model. You have an employee, they have a username and password (and maybe an MFA code), and the system verifies they are who they claim to be.

AI agents break this model completely.

An AI agent does not have a keyboard. It does not receive a text message with a one-time code. It may spin up hundreds of instances in seconds, access sensitive systems across multiple environments, and make decisions at a speed and scale no human approver can match. It interacts with data, tools, and APIs continuously, autonomously, and with permissions that were granted once but never reviewed again.

NewCore’s argument — backed by serious investors — is that existing platforms treat AI agents as an afterthought, bolting on access credentials that were fundamentally designed for a different world.

What NewCore Actually Builds

NewCore’s platform is built from the ground up for what the company calls the “agentic era” of enterprise work: environments where AI agents and human workers operate side by side, accessing the same systems and data.

The core of the platform includes:

Split-key signing architecture. Rather than storing a single credential that could be compromised, NewCore splits signing authority across multiple components. No single point of failure can compromise an agent’s identity.

Hardware-bound credentials. Agent identities are tied to hardware roots of trust, making them far harder to steal or spoof than software-based tokens.

Phishing-resistant authentication. Since agents cannot be phished the traditional way, NewCore’s approach focuses on cryptographic attestation rather than knowledge-based factors.

Continuous identity verification. Unlike human sessions that authenticate once at login, agents operating over extended periods re-verify continuously as they move between systems and tasks.

The company is led by Zohar Alon, who previously co-founded Dome9 Security, acquired by Check Point in 2018. NewCore is being demonstrated this week at Identiverse 2026 in Las Vegas.

Why This Is Happening Now

The timing is not accidental. Enterprise AI agent deployments have accelerated dramatically in 2025 and 2026. Businesses are deploying agents to handle customer interactions, process documents, run internal workflows, and make operational decisions that previously required human sign-off.

As these agents gain access to more systems and more sensitive data, the security questions intensify. Earlier this year, research from security firms found that AI agents at major enterprises were frequently operating with over-provisioned access — permissions that were granted broadly and never scoped to what the agent actually needed. Some agents were found accessing systems outside their intended scope simply because no one had set up the governance infrastructure to prevent it.

The identity layer is where that governance has to start. You cannot audit what an agent did if you cannot reliably know which agent did it. You cannot enforce least-privilege access if every agent instance looks the same to your security tooling.

What This Means for Business

If your business is deploying AI agents — or planning to — this story is a signal worth paying attention to.

The compliance exposure is real. Regulators across the US, EU, and Australia are actively developing frameworks that require businesses to account for automated systems accessing personal data or making consequential decisions. “An AI agent did it” is not an answer that satisfies a data breach investigation or an audit. Businesses need to know exactly which agent accessed what, when, and with whose authorisation.

Governance needs to be designed in, not bolted on. The pattern with every prior wave of enterprise technology — cloud, mobile, SaaS — was that security came last, was expensive to retrofit, and created risk in the meantime. AI agents are following the same trajectory. The businesses that establish identity and governance infrastructure now will avoid the scramble later.

Not every business needs a platform like NewCore yet. If you are running a small number of agents on a contained set of tasks, your existing IAM tooling can probably manage. But if you are building toward an AI-native operation — multiple agents, broad system access, autonomous decision-making — the question of how you will verify, audit, and control those agents deserves a concrete answer before you scale.

The market signal matters. When seasoned cybersecurity investors back a $300 million seed-stage company to solve a specific enterprise problem, it is confirmation that the problem is real and large. Cyberstarts in particular has a track record of backing companies that address enterprise security gaps before the breach forces the issue.

Enterprise DNA has been watching the identity and governance layer of agentic AI closely. The businesses succeeding with AI agent deployments are not the ones moving fastest — they are the ones building with control built in from the start. NewCore’s emergence is a reminder that deploying agents is only the beginning of the work.

If you are at the stage of figuring out where AI fits in your business, a discovery conversation is the right place to start. We talk through what you are building toward, where the real risks are, and what governance infrastructure you will need to support it.