Abnormal Security vs Anthropic

Discovers shadow AI tool/agent/chat usage via email, OAuth, identity and browser signals and enforces policy automatically.

Answers security teams' questions about threats, attack patterns and risk across email and SaaS on demand, without writing queries.

Phishing simulation and training that generates its exercises from the real attacks aimed at that organisation rather than from static templates.

Automates phishing triage: answers every employee report in seconds and remediates campaign-wide threats without an analyst in the loop.

Combines a custom outbound DLP policy engine with an AI triage agent that evaluates matches in context, releasing benign messages and quarantining likely violations.

Behavioral AI that automatically moves low-priority promotional email (graymail) out of the inbox without manual rules.

Behavioral AI that catches hijacked sessions, abused OAuth grants and privilege misuse across email, IdP and SaaS.

Behavioral-AI email threat detection that builds per-employee/vendor baselines to catch BEC, phishing and payload-free social engineering before delivery.

Analyzes identity signals pre-provisioning (HR systems, IdP, email) to detect synthetic personas and coordinated nation-state infiltration campaigns.

Inspects Microsoft Teams and Slack for malicious URLs and weaponized attachments, auto-remediating high-risk messages.

Uses behavioral AI to detect outbound email sent to unintended recipients and quarantines it before delivery.

Checks a Microsoft 365 environment against CIS Benchmarks, detects configuration drift in real time and provides prioritized remediation.

Detects and autonomously remediates compromised Microsoft 365/Google Workspace accounts by learning normal sign-in, device and behavioral patterns.

Assistant for chat, document analysis and coding, used in the browser and in desktop and mobile apps.

Lets users collaborate with Claude to create on-brand visual work (decks, prototypes, wireframes, marketing collateral) with design-system import and Claude Code integration via /design-sync.

Desktop application for scientific research that runs analyses, searches scientific databases and traces work from data wrangling through to publication.

Application that scans codebases for high-severity vulnerabilities, validates findings by adversarial verification and recommends patches for review, with a companion Claude Code plugin.

Claude working inside Excel, PowerPoint, Word and Outlook with context carried across the documents — preserving templates and house styles and showing every edit as tracked changes, highlighted cells or draft emails.

Claude tagged as @Claude inside Slack threads and channels: it reads thread context and replies in real time, analyses conversations, pulls data, drafts content, monitors channels, triages alerts and runs standing instructions without leaving Slack.

Coding agent that works on a codebase from the terminal, an IDE extension and a desktop app.

Agentic task app in which Claude works asynchronously across a user's files and tools toward a stated goal, checked in on from desktop, web or phone.

Browser extension that lets Claude act inside Chrome on the user's behalf — reading pages, clicking, typing and filling forms to work through multi-step web tasks without approving each step, with the user keeping control of what it may do.