Cogent

Today, we’re launching Cogent VR-1, our Mythos-class cyber reasoning model built for the enterprise. Frontier models have shown they can discover vulnerabilities in code. VR-1 is designed for the harder problem: reasoning across the messy reality of a live enterprise. Cloud. Identity. Applications. Endpoints. Data. Misconfigurations that look harmless in isolation, but become critical when chained together. VR-1 can autonomously map those connections, build working attack paths, show how an attacker could move through an environment, and verify whether a fix actually closes the gap. On IntrusionBench, our new execution-verified benchmark for enterprise attack chains, VR-1 completed twice as many hidden attack paths as other frontier models. This is the capability attackers are rapidly gaining. Defenders should benefit from it too, grounded in their environment, constrained by clear rules of engagement, and deployed through a governed security harness. VR-1 is available through the Cogent Frontier Access Program. Learn more: cogent.com/frontier

We're unveiling a new Mythos-class frontier model built for cybersecurity, live on July 27 at 10am PT / 1pm ET. Frontier models are already good at finding an isolated bug in a single codebase. Real enterprise attacks rarely stay that clean. We built a model for that problem. Give it a single foothold inside an enterprise and it reasons across the environment the way an attacker would, then finds the exact path to the data that matters most and how to close it.

Today we're announcing Autonomous Vulnerability Response (AVR), the first end-to-end autonomous AI vulnerability response capability proven in production. Most vulnerability management tools stop at the ticket. AVR runs the whole loop as one continuous process: 1. Detection: agentless, catches zero day exposures within ~10 minutes of disclosure. 2. Investigation: gathers evidence across the environment, correlates against asset ownership and compensating controls, produces a findings package with reasoning preserved. 3. Remediation: generates the fix, scores change impact, dispatches through Jira or ServiceNow, and executes under customer guardrails. 4. Validation: independently confirms the fix through a scanner rescan or SBOM rebuild. If it didn't hold, the ticket reopens automatically. Where customer policy allows, the loop runs without human intervention. Where policy requires oversight, humans stay in. The video below is with Corey Kaemming (CISO, Valvoline Instant Oil Change) and Shawn Kerns (Head of InfoSec, University of the Pacific), two of the practitioners running AVR in production. Full write-up in the comments.

Cogent’s AI agents handle the tedious work that happens after vulnerabilities are detected, so you can close critical risks faster.

Cogent’s AI agents handle the tedious work that happens after vulnerabilities are detected, so you can close critical risks faster.

Cogent’s AI agents handle the tedious work that happens after vulnerabilities are detected, so you can close critical risks faster.