Cogent and OpenAI Launch AI Agents for Enterprise Vulnerability Fixes

Cogent and OpenAI have launched a system that uses AI agents to help enterprise security teams investigate software vulnerabilities and verify fixes. The agents can assess enterprise exposures, make permitted changes and verify whether the risks have been addressed.

Cogent’s Context Engine maintains a shared map of company assets, vulnerabilities, identities, owners and dependencies. That context helps agents work across security tools. The initiative integrates GPT-6.1 Sol and GPT-5.6; GPT-6 Astra supports upfront planning and complex reasoning.

Customers set limits on what the agents can do. Cogent’s TRACE engine checks tool calls against company policies, while its Agent Gateway separately checks authorization. Higher-risk tasks run in isolated sandboxes. Customers can require approval for changes or allow a defined set of remediations to proceed autonomously.

Cogent says the system cut remediation time by 75% and increased the number of vulnerabilities triaged fivefold.

“AI agents can now carry work from investigation through verified remediation,” said Geng Sng, Cogent’s co-founder and CTO. Cogent uses its Cyber and Daybreak platforms to study attack paths and defensive techniques.

In a separate effort, OpenAI has urged Congress to fund and authorize a Cybersecurity and Infrastructure Security Agency-led cyber defense plan for government networks and critical infrastructure.