Cogent Security has unveiled Cogent VR-1, a frontier AI reasoning model purpose-built for cybersecurity and designed to operate within live enterprise environments. Unlike conventional AI security models that focus on isolated vulnerabilities, VR-1 is trained to identify and validate complex, multi-step attack paths across enterprise infrastructure. Alongside the launch, the company introduced IntrusionBench, a benchmark that evaluates AI models on realistic enterprise attack scenarios, where VR-1 achieved twice the performance of competing frontier AI models.
Most AI models today excel at identifying isolated vulnerabilities within individual codebases. However, enterprise cyberattacks typically involve a sequence of interconnected weaknesses that span multiple systems. A seemingly minor software flaw, excessive identity permissions, or an unverified build artifact may appear harmless individually but can collectively create a pathway to critical enterprise assets.
Cogent trained VR-1 to operate like an experienced security researcher by mapping enterprise environments, connecting vulnerabilities across systems, validating attack paths, and verifying whether remediation efforts eliminate the risk rather than simply shifting it elsewhere. According to the company, tasks that previously required weeks of manual security analysis can now be completed in hours.
The announcement comes as autonomous AI-powered cyberattacks increasingly target governments and major technology organizations, creating demand for equally advanced AI-powered defensive capabilities.
"For two years, every security vendor has claimed to be AI-native. VR-1 lets us put a number on it," said Vineet Edupuganti, CEO of Cogent Security. "This is Mythos-level security reasoning, running inside a customer's actual environment. We can show them exactly which attack paths a frontier-capable adversary could reach, then use the same model to close them. Attackers already have this capability. Our job is to make sure defenders get it first, and that it runs under real controls."
To demonstrate VR-1's capabilities, Cogent introduced IntrusionBench, a cybersecurity benchmark designed to simulate realistic enterprise breaches rather than isolated software exploits.
Instead of testing whether an AI model can exploit a single vulnerability, IntrusionBench evaluates whether it can navigate cloud infrastructure, identity systems, and internal enterprise tools after gaining an initial foothold. Success is measured by executed attack paths rather than theoretical reasoning.
According to Cogent, VR-1 significantly outperformed Kimi K3, Opus 4.8, and GLM-5.2 in the benchmark's most challenging configuration, where the AI agent had no prior knowledge of the target environment. As additional environmental information became available, competing models improved, suggesting VR-1's primary advantage lies in investigating uncertainty rather than exploiting known vulnerabilities.
"The frontier models security teams use today can find an attack path, but only as a byproduct of being good at code and reasoning. No one set out to build a model for this, so we did," said Geng Sng, co-founder and CTO of Cogent Security. "VR-1 is trained specifically for the multi-step attacks that play out in real, messy enterprise environments. That focus is what moves it past the general-purpose frontier models on the problems defenders actually face."
Alongside VR-1, Cogent introduced the Cogent AI Harness, a runtime environment designed to safely deploy AI security models within enterprise environments.
The AI Harness provides enterprise context, scoped tools, policy enforcement, and evidence-backed verification for AI-driven security operations. Every AI action is validated against customer-defined security policies while following Cogent's "prove, don't detonate" principle, confirming security exposure without causing operational disruption or damaging enterprise systems.
The platform supports both frontier and open-weight AI models while maintaining governance controls suitable for enterprise security teams.
Recognizing the dual-use nature of advanced cybersecurity AI, Cogent is restricting VR-1 availability through its Cogent Frontier Access Program instead of releasing it publicly.
Only vetted organizations will receive access to the model, along with built-in safety controls, policy enforcement, and audit logging. Participating organizations can also work directly with Cogent Research to evaluate the model within their own environments.
Eligible participants can receive a Frontier Model Risk Assessment that identifies potential AI-enabled attack paths, highlights vulnerabilities that become more critical as AI capabilities evolve, and recommends areas where AI can safely strengthen enterprise security.
Cogent's launch reflects the growing emphasis on specialized AI reasoning models that help enterprises proactively defend against increasingly sophisticated AI-powered cyber threats while maintaining strict operational governance.
Cogent is an applied AI lab whose agents detect and fix security vulnerabilities faster than attackers can exploit them. The Cogent platform identifies exposure to new vulnerabilities within minutes, builds contextualized remediation plans, and executes fixes at whatever level of autonomy the customer allows, from human-approved to fully autonomous. Fortune 500 security teams using Cogent have reduced the exposure window for critical vulnerabilities by 97 percent. Built by researchers and operators from Google DeepMind, Abnormal Security, and Coinbase, Cogent is backed by Greylock Partners and Bain Capital Ventures.
TechIntelPro empowers B2B technology decision-makers with the latest industry trends, deep insights, leadership perspectives and actionable intelligence to drive both organizational and professional milestones, and thrive on the tech pulse.
