Abnormal AI has expanded its behavioural security platform with three new products designed to protect identity systems, AI tools, and recruitment pipelines from attacks that appear legitimate.
The company has launched Identity Threat Protection, AI Governance, and Infiltration Prevention, alongside an AI App Store that allows customers to activate Abnormal products from a central interface. The products go live on 3 August 2026.
The expansion takes Abnormal further beyond email security and into the wider identity attack surface. The company said its behavioural AI foundation, already used to protect more than 4,500 customers, will now be applied to authenticated sessions, AI adoption, service accounts, OAuth apps, AI agents, and suspicious candidate identities.
“Abnormal envisioned a future where behavioural AI could solve critical problems facing enterprises around the world,” said Evan Reiser, CEO and co-founder of Abnormal AI. “Our customers are facing key challenges today: attackers who blend in instead of breaking in, AI adoption outpacing governance, and nation-state actors engineering fake identities to become insiders. Every one of those risks comes down to understanding identity and behaviour. That’s the lens that Abnormal brings to these problems.”
Identity Threat Protection is intended to identify weaknesses such as accounts without multi-factor authentication and overprivileged service accounts, then rank them by likely attack susceptibility. It also maps real-world attack techniques, including adversary-in-the-middle phishing and OAuth abuse, to each customer’s environment.
AI Governance is designed to discover, score, and govern AI tools, agents, and chats across an organisation, including sanctioned and unsanctioned use. Abnormal said the product establishes behavioural baselines for each tool and agent, flags deviations, and applies customer-configured policies automatically when activity appears risky.
Infiltration Prevention focuses on identifying potentially fraudulent candidates before enterprise access is provisioned. It integrates with applicant tracking systems including Greenhouse and Workday, using signals such as VoIP burner numbers, VPN-masked locations, and threat intelligence linking repeated suspicious behaviour.
The release reflects a wider change in security priorities, with identity, behaviour, and governance gaining prominence alongside perimeter controls. Attackers increasingly seek to operate inside valid sessions, use legitimate credentials, exploit workflow trust, or enter organisations through supplier, help desk, or hiring processes. Traditional defences that check credentials at login may not detect what happens once an identity has been accepted.
The launch sits alongside a wider enterprise shift towards governed automation, with governed agents being introduced into core operations and cyber resilience obligations rising in the UK. AI systems, identities, and automated workflows are becoming part of the attack surface and part of the control environment.
The AI Governance product is especially relevant as organisations adopt AI faster than many security and compliance teams can inventory it. Employees may use public tools, embedded assistants, workflow agents, code generators, and SaaS features without a complete central record. That creates exposure around data leakage, excessive permissions, shadow AI, and unclear accountability.
Non-human identities add another layer. Service accounts, bots, integrations, and AI agents can outnumber employees, often with access to systems and data that would be tightly controlled for a human user. If those identities are poorly governed, they can become privileged pathways for attackers.
“The threats that matter most don’t look like threats at all,” Reiser added. “A trusted identity moves through a valid session like a real employee. A helpful AI agent codes its way into a sensitive database. A nation-state actor uses a synthetic persona to apply for a job. Each looks different, but each gives itself away the same way, through abnormal behaviour.”
The AI App Store also points to a commercial change in enterprise security. Vendors are trying to make adoption faster as threats change, but rapid activation must still be balanced with procurement control, risk assessment, integration testing, and auditability.
Abnormal’s expansion shows how the security market is reorganising around behaviour rather than only credentials, rules, or signatures. As AI agents and digital identities gain authority across the enterprise, security teams will need to see when a trusted identity starts acting in a way that no longer deserves trust.
You must be logged in to post a comment.