Oracle announced an AI security service designed to help enterprise teams investigate vulnerabilities across their systems, with customer approval required before it applies remediation changes.
The company introduced Oracle AI Security Mesh on Oct. 27. It combines OpenAI cyber models with Oracle testing tools and information customers supply about the systems under review. Oracle says the service is intended to cover on-premises deployments, Oracle Cloud Infrastructure (OCI) and other cloud environments.
The models are designed to trace ways a weakness could be used in an attack and help security teams rank issues for attention. Oracle says the service can provide steps to reproduce a vulnerability and testing findings. It can also run proposed repairs through tests to check whether applications behave as expected.
Oracle said customers must approve a remediation change before the service puts it into effect. The company says its AI agents receive individual identities and permissions for their assigned tasks. Separate monitoring tools can detect or stop actions that go beyond the limits approved for a task.
OpenAI said Oracle operates the testing environment, a design intended to help customers keep use within the scope they have authorized. The announcements did not identify the specific OpenAI models involved or provide benchmark results for detecting or fixing vulnerabilities.
Oracle also highlighted two related controls in Oracle Cloud Infrastructure. OCI Identity Assurance relies on biometric checks to confirm a person's identity, while OCI Zero Trust Packet Routing governs which network connections are allowed. Oracle said it launched the routing service in 2024.
Phil Newman, Oracle's senior vice president of Security and Identity, said, “As AI agents take on more responsibility, enterprises need security that can keep pace with how these systems operate.”
Oracle did not give a general availability date or price, and its announcement said release timing and pricing could change.
