The Security Model Broke When Agents Started Working
AriaCymphony just raised $30M to fix identity governance for a world where your coworkers are algorithms.
The IAM Gap
Your identity access management system was built for humans. It assumes a person logs in, does their job, and logs off. Sessions last hours. Permissions change quarterly. Audits happen monthly. When someone leaves the company, an admin revokes their access and moves on.
None of that holds when an autonomous agent executes a thousand API calls in thirty seconds.
Traditional IAM tools authenticate users. They do not authenticate intent. They verify that you are who you claim to be. They do not verify whether the database write you are attempting at 3 AM serves a legitimate business purpose or represents a cascading failure in an agent's decision tree. They track login times and session durations. They do not track whether an agent that started a customer onboarding workflow has somehow migrated into your financial reporting system.
This is the gap Cymphony stepped into.
On September 9, 2026, the cybersecurity startup emerged from stealth with $30 million in Series A funding co-led by Sequoia Capital and SMBC Fin Atlas Beyond Fund aipressroom.com. The founders, Shy Dekel, Idan Berkovits, and Edi Gotlieb, spent years in enterprise security before concluding that the next major breach would not start with a compromised human credential. It would start with an over-privileged agent token that nobody noticed because nobody was looking for it.
Workforce Graph: One Map for Humans and Agents
Cymphony's core product is a workforce graph that maps human employees, AI agents, and non-human identities into a single permission structure. Think of it as an org chart that includes every service account, every bot, and every autonomous workflow alongside the people who manage them.
The graph tracks who has access to what, which agents interact with which data stores, and where permission boundaries blur. When a marketing agent needs to pull customer records from your CRM, the graph knows. When that same agent suddenly starts querying your financial database, the graph flags it. When a new agent spins up without anyone registering it in the system, the graph detects the anomaly.
This visibility matters because most enterprises do not know how many AI agents they are running. A 2025 Gartner survey found that 60% of organizations deploying autonomous agents could not accurately count them gartner.com. Shadow agents proliferate when development teams spin up autonomous workflows without going through security review. You cannot secure what you cannot see.
The workforce graph also maps data interaction patterns. It knows that your sales agent typically reads from the CRM and writes to the email platform. It knows that your data pipeline agent reads from the data warehouse and writes to the analytics dashboard. When either agent deviates from its established pattern, the system notices.
Automated Remediation at Machine Speed
Visibility without response is just a dashboard. Cymphony pairs its workforce graph with automated remediation that isolates over-privileged agent tokens before rogue behaviors cascade through enterprise pipelines.
The system uses specialized AI agents to investigate access anomalies. When an agent's behavior deviates from its established pattern, the remediation engine can revoke permissions, quarantine the token, or block the action entirely. No human ticket. No waiting for a security analyst to review the alert. No delay between detection and response.
This speed is not a luxury. It is a requirement. When an agent can execute a destructive database operation in the time it takes a human to read an email, manual review is not a safeguard. It is a post-mortem. The remediation has to happen at machine speed because the threat moves at machine speed.
The automated agents that handle remediation are themselves governed by the workforce graph. They operate within strict permission boundaries. They can revoke tokens, but they cannot escalate their own privileges. They can quarantine agents, but they cannot access the data those agents were touching. The system is designed so that the security agents themselves cannot become the threat.
Why Sequoia Bet $30M on Agentic Security
Sequoia's investment signals something specific: agentic security is not a feature of existing IAM platforms. It is a distinct category with its own requirements, its own architecture, and its own market dynamics.
The distinction matters because the threat model changes when the actor is non-human. Human attackers need to compromise credentials through phishing, social engineering, or brute force. Over-privileged agents already have valid credentials issued by the organization. The question is not whether they are authenticated. The question is whether their current task justifies their current access level.
Traditional security tools answer the first question. Cymphony answers the second.
The market timing is significant. Enterprise AI agent deployments are accelerating. McKinsey estimates that autonomous agent workflows will handle 30% of all enterprise software interactions by 2028 mckinsey.com. Each of those interactions is a potential security event. The current tools were not built for this volume or this velocity.
The Category Split
Enterprise security budgets are pivoting. For two decades, identity governance meant managing human access. The tools, the vendors, the compliance frameworks all assumed a human at the keyboard making intentional decisions.
That assumption is now wrong.
As agentic workflows replace basic copilots, the security perimeter shifts from the login screen to the action layer. Every database write, every API call, every file access becomes a policy decision that must be evaluated in real time. The policy engine has to understand not just who is making the request, but what they are trying to do and whether that action aligns with their established role.
Cymphony is betting that enterprises will spend billions securing this new layer. Sequoia's $30M says they agree. The agent era did not arrive with a research paper. It arrived with a funding round for the security model that should have existed before the agents did.