About Abnormal AI
Abnormal AI sells a behavioral security platform that protects email, identity, and AI. Abnormal says AI-generated attacks are often too subtle for humans to see and too new for traditional threat intel to detect. Abnormal AI says its platform stops the attacks that slip past legacy gateways, including business email compromise, advanced phishing, vendor fraud, and account takeover. Abnormal AI states its mission is to stop crime with AI.[15]
Abnormal AI says its platform connects AI tools, people, agents, and cloud activity through behavioral context to govern adoption, enforce policies, detect threats, and support investigations. The platform has native API integrations across Microsoft 365, Google Workspace, Okta, CrowdStrike, and the customer's SIEM and SOAR. Abnormal AI says each supported connection helps Abnormal understand activity, investigate threats, and apply the controls the customer's team authorizes. Abnormal AI describes Attune as the behavioral AI model that learns what is normal and detects what does not fit. AI Data Analyst lets users ask plain-English questions and get security insights instantly. AI Data Analyst consolidates insights from across the Abnormal environment into a single source of truth.[15][16][20]
For email, Abnormal AI describes Inbound Email Security as stopping BEC, phishing, and credential theft before delivery. AI Security Mailbox is described as an AI co-worker for email security operations. AI Phishing Coach is described as just-in-time training that adapts to each user's real risk. Messaging Security is described as detecting and remediating malicious content in messaging apps, and Misdirected Email as preventing accidental data leakage from misaddressed email. Email DLP Rules combines a custom policy engine with an AI Triage Agent. Teams can launch Email DLP Rules in passive mode, logging matches without taking action, then move to active enforcement. Email Productivity uses behavioral AI to move low-priority promotional email out of the inbox, and Email Productivity deploys through API integration.[15][17][18][21]
For identity, Account Takeover Protection detects and remediates compromised Microsoft 365 and Google Workspace accounts by learning normal sign-in, device, and behavioral patterns for each user. Account Takeover Protection locks attackers out by revoking active sessions and forcing credential resets. Abnormal AI says Account Takeover Protection connects via API to Microsoft 365 and Google Workspace. Abnormal AI says brute force, credential stuffing, and token or session theft produce sign-ins that satisfy authentication. Abnormal AI says its platform correlates email, IdP, and SaaS signals into one behavioral model, catching privilege misuse and socially-engineered MFA resets. Identity Threat Protection correlates signals to contain threats before and after authentication. Infiltration Prevention is described as detecting synthetic personas and infiltration campaigns before access is provisioned. Posture Management is described as catching risky misconfigurations across the email platform.[3][15][19]
AI Cloud Security is described as detecting AI-driven attacks against cloud identities and workloads and containing threats within the customer's response policies. AI Cloud Security is at the Priority First stage, available for purchase and production deployment to selected customers. AI Employee Guardrails is described as keeping employee AI use on task, protecting sensitive data, and enforcing policies based on the person, tool, and request. AI Agent Security is described as inventorying production agents and verifying that each one behaves as intended. AI Agent Security is at the Design Partner stage, where selected customers evaluate the product and give regular feedback to the product team. AI Employee Guardrails is at the Design Partner stage. AI Security Workbench is described as running AI-powered investigations across petabytes of behavioral events.[3][15][20][21][22]
Abnormal AI says it powers 5,000 customers, including 30% of the Fortune 500.[16][17][18][19][20][21][22]
Abnormal AI on the record