# Offroad Launches With $7M to Combat Enterprise Identity Crisis: AI Agents Now the Target
New startup uses autonomous security agents to find, govern, and remediate identity risks across sprawling enterprise systems
Identity security is entering a new phase of complexity. As enterprises grapple with exploding numbers of human users, machine identities, AI agents, and third-party applications, traditional identity management approaches are breaking down under the load. Offroad, a New York- and Tel Aviv-based startup that emerged from stealth this week, is betting that the only way forward is to fight complexity with complexity—deploying autonomous AI agents to hunt down and resolve identity risks that human teams can no longer track manually.
The company announced $7 million in seed funding, led by Ibex Investors and Skywell Capital, to scale its mission of moving enterprises from basic identity visibility to full identity resolution. Co-founded in May 2025 by CEO Dan Bendler and CTO Philip Shteyn (formerly a Captain at Unit 8200, Israeli Military Intelligence), Offroad targets a problem that has been simmering for years but is now reaching a critical inflection point: the identity landscape has become so fragmented and rapidly changing that security teams are drowning in risk they cannot hope to remediate.
"Enterprises now operate across a constantly changing mix of human users, machine identities, and AI agents," Bendler said in an interview. "The context needed to understand and resolve identity risk is spread across dozens of systems and workflows, while security teams are still expected to investigate and remediate issues manually. That model is becoming increasingly difficult to sustain."
## The Threat: Identity Sprawl in an Age of Autonomous Systems
The fundamental problem Offroad is targeting reflects a seismic shift in how enterprises operate. Five years ago, the identity landscape was simpler: you had user accounts, some service accounts, and a manageable number of integrations. Today, the picture is radically different.
The identity explosion includes:
The cumulative effect is an identity landscape that has become fundamentally unmanageable through manual processes. Security teams face a choice between accepting risk they cannot properly investigate or burning enormous resources on remediation that will never catch up to the growth rate of identities themselves.
"Most identity systems were designed around assumptions that no longer hold," said Shteyn. "AI agents operate across systems at all hours and at a scale humans never could, which makes traditional behavioral baselines far less reliable. Security teams need systems capable of continuously investigating and reasoning through identity activity, not simply generating more findings."
## Background and Context: Why Identity Risks Are Exploding Now
The timing of Offroad's launch is significant. Enterprise identity risk has been building for years through several overlapping trends:
1. The AI Agent Revolution
The rapid deployment of AI agents to automate business processes has introduced a new identity category that traditional access controls were never designed for. Unlike human users who follow relatively predictable patterns, AI agents operate 24/7, access multiple systems simultaneously, and make decisions at machine speed. Traditional baselines for "normal" behavior break down when the actor is an autonomous system.
2. API-First Architectures
Modern cloud-native development relies heavily on APIs and integrations. Each integration requires credentials and permissions. Each permission represents a potential attack surface. Organizations that have deployed dozens or hundreds of cloud applications rarely have comprehensive visibility into what permissions have been granted and who or what is using them.
3. Third-Party Application Proliferation
Employees increasingly request access to specialized SaaS tools for specific functions. Each tool typically requests OAuth scopes or API permissions. Over time, these permissions are rarely revoked even when the tool is no longer used or when employees change roles.
4. Credential-Based Attacks and Lateral Movement
Recent major breaches have demonstrated that stolen credentials—whether from compromised users or breached third-party applications—are among the most effective attack vectors for lateral movement and persistent access. Identity is no longer simply an access control problem; it is a critical security perimeter.
## Technical Details: How Offroad Approaches Identity Risk
Offroad's solution centers on using autonomous AI agents to handle the investigation, governance, and remediation of identity risks that have become too numerous and complex for manual handling.
The company's approach involves three layers:
| Layer | Function | Example |
|-------|----------|---------|
| Investigation | Continuous monitoring of identity behavior and configuration; automatic context gathering across fragmented systems | Identifying that a service account has permissions far exceeding its stated function |
| Governance | Using agentic AI to reason through whether identified risks warrant remediation | Determining if an over-privileged account is genuinely needed or represents legacy access that should be revoked |
| Remediation | Either reporting findings to human analysts or—where safe—executing remediation autonomously | Automatically removing unnecessary permissions or revoking dormant credentials after human review |
The key innovation is the "human in the loop where necessary" approach. Not all remediation is automated. Instead, Offroad's agents determine which findings require human judgment and which can be safely remediated autonomously.
### The OAuth Audit Report: Revealing Hidden Risk
To demonstrate the scale of the problem, Offroad published a detailed audit of 2,890 public OAuth applications available on the Google Workspace Marketplace and GitHub Marketplace. The findings are sobering:
- Scopes wider than the app's stated function
- AI components with write access to organizational data
- Threat-intelligence flags indicating suspicious publisher behavior
- Dead publisher websites indicating abandoned or compromised apps
- Buyable or pending domain names (indicating domain squatting risk)
- Brand-leading app names published by unknown or third-party entities
These findings suggest that one in three publicly available OAuth applications poses a structural security risk. For organizations that have deployed dozens of these applications across their Workspace or GitHub environments, the cumulative risk is substantial.
### ohauth.ai: Community-Driven Risk Cataloging
In parallel with the audit report, Offroad launched ohauth.ai, described as "a community catalog of OAuth apps with over-privileged scopes, dead publisher domains, and silent permission drift." This initiative transforms the findings from a one-time audit into an ongoing, crowdsourced resource that the security community can contribute to and benefit from.
## Implications for Organizations
Offroad's emergence highlights several critical implications for enterprise security teams:
1. Manual Identity Management Is No Longer Viable
Organizations that rely on quarterly access reviews and manual provisioning workflows are increasingly exposed. The velocity of change in modern enterprises outpaces the speed at which humans can investigate and remediate.
2. Third-Party Applications Require Continuous Scrutiny
The OAuth audit findings suggest that off-the-shelf applications—particularly those offering integration with core enterprise platforms—should be treated as potential security liabilities until proven otherwise. Vetting should include publisher legitimacy, scope justification, and ongoing monitoring for permission drift.
3. AI Agents Require New Security Primitives
Traditional identity and access management controls were designed for human users and long-lived service accounts. AI agents that operate across multiple systems, at high velocity, and with autonomous decision-making capability require new approaches to risk assessment and containment.
4. Identity Risks Build Quietly Over Time
The company notes that while some identity risks manifest immediately through suspicious behavior, others "build quietly over time through privileges that outlive their purpose, access that's kept after roles change, [and] third-party apps with permissions nobody can justify." Continuous monitoring and periodic re-evaluation of all granted permissions are essential.
## Recommendations for Defenders
Organizations seeking to improve their identity risk posture should consider the following:
---
## HackWire Analysis
Offroad's $7 million funding round and the scale of findings in their OAuth audit reflect a fundamental shift in how enterprise security teams must approach identity risk. The critical insight here is not simply that identity management has become complex—it's that human-speed remediation is no longer compatible with machine-speed threats.
The traditional security model assumes that after a vulnerability or misconfiguration is discovered, a human analyst will investigate, validate, and remediate it. This works when discoveries are infrequent. It fails catastrophically when autonomous systems can discover and exploit misconfigurations faster than humans can review them.
What's particularly revealing about Offroad's OAuth audit is not just the 32% figure, but what it represents: visible risk in the public ecosystem. If one-third of publicly available OAuth applications have structural weaknesses, the proportion of internal, custom-built integrations with permission sprawl is likely far higher. Most organizations have never conducted a systematic audit of all OAuth or API permissions granted across their entire footprint.
The AI agent dimension adds urgency. Unlike human users, AI agents can traverse your identity landscape in seconds, follow dozens of permission paths simultaneously, and operate in ways that don't trigger traditional behavioral anomalies (because unusual behavior from an AI is normal behavior). A compromised AI agent with broad permissions is not just a data loss risk; it's a lateral movement superhighway.
Offroad's positioning of autonomous remediation—human in the loop where necessary—is clever because it sidesteps the false choice between "zero trust" (impractical at scale) and "trust but verify" (too slow). Instead: continuous investigation, threshold-based automation, and human judgment reserved for ambiguous cases. Whether Offroad executes on this vision is an open question, but the direction is sound.
The broader trend this signals: identity risk is becoming an endemic feature of modern enterprise architecture, and the only tools that can manage it at scale will themselves need to be autonomous. Security teams will increasingly delegate identity governance to AI agents precisely because the problem has become too large for human cognition.
— HackWire Editorial
---
## Related Coverage