Abnormal AI expands security suite with five products

Abnormal AI has expanded its AI Security suite with three new products and the general release of AI Governance. The package also includes AI Cloud Security, which remains in private preview.

The move broadens the company's security offering as businesses increase their use of generative AI tools and software agents in everyday work.

The updated suite now spans five products covering different parts of AI oversight and threat detection. Alongside AI Governance and AI Cloud Security, the new additions are AI Employee Guardrails, AI Agent Security and AI Security Workbench.

The products are built on Abnormal AI's behavioural security platform, which analyses activity across users, identities and systems. The suite is designed to give security teams more visibility into which AI tools are in use, who or what can access sensitive systems and data, and whether behaviour departs from expected patterns.

Five areas

AI Governance is the first product to reach general availability. It is designed to help organisations discover both approved and unapproved AI applications in use across the business, assess adoption, set policies, track spending and retain records for governance, audit and compliance work.

AI Cloud Security focuses on cloud environments. It is intended to identify unmanaged or potentially malicious AI activity, show which identities and resources are involved, surface security and configuration risks, and support investigations across production cloud systems.

AI Employee Guardrails is aimed at staff using generative AI tools and AI-enabled applications. According to the company, security teams can apply policies at the point of use and either guide or block activity that creates security or data risks.

AI Agent Security targets both first-party and third-party AI agents. The product links agents to identities, permissions, systems and resources, then monitors their actions for unexpected or risky behaviour.

AI Security Workbench is positioned as an investigation environment for security teams. It allows engineers to analyse activity across behavioural security data, identify patterns worth examining and develop detections and response workflows without relying entirely on manually written queries and rules.

Wider shift

The launch reflects a broader change in enterprise security priorities as companies move from experimenting with AI tools to embedding them in routine operations. That shift has created a parallel problem for security teams, which now need to manage both the use of AI inside the business and the growing use of AI by attackers.

Abnormal AI framed the issue around identity, access and behaviour across both human and non-human actors. That includes employees using public or approved AI tools, software agents acting with delegated permissions, and AI-related activity inside cloud infrastructure.

"Advances in AI are changing the risk profile for organizations around the world," said Evan Reiser, Chief Executive Officer and Founder of Abnormal AI. "That makes understanding human and non-human identity and behavior more important. Security teams need to understand who is acting, what access they have, what normal behavior looks like, and when something meaningfully changes. That is the security problem Abnormal was built to solve, and we are now extending that approach to enterprise AI. Enterprises need one platform that sees and understands behavior for employees, agents, and cloud infrastructure, because that's how these risks actually surface."

The suite is intended to address both governance and defence. On one side, organisations want clearer oversight of which AI services are being used, whether those services are authorised and how they interact with sensitive information. On the other, security teams are increasingly concerned about threats involving AI-generated attacks, autonomous agents and abnormal cloud activity.

Existing customers can activate AI Governance through Abnormal AI's app store. The other four products remain in private preview.

Abnormal AI says it protects more than 5,000 organisations, including 30% of the Fortune 500. The company has built its market position around behavioural analysis, arguing that security teams need to judge actions not only by permissions but also by whether activity matches expected behaviour.

Reiser said the same principle applies to AI use inside companies. "Security teams should not have to choose between enabling AI and controlling the risk that comes with it," he said. "The goal is to give them enough visibility and context to let the business move quickly, while recognizing when an employee, an account, or an autonomous agent starts doing something it should not be doing."