Proactive Insider Risk
for Humans and AI Agents

Above's AI investigators work every insider and AI-agent risk to a verdict and hand your security team the case. Real-time guidance stops the next one before it starts.

A wide isometric city of pale blocks reaching back to the foot of pale mountains, with downtown towers, civic halls, a station and warehouses, a river that runs on as tree-lined canals across the whole city under small bridges, small parks with fountains and paths, and streets. At the right an ivory griffin sits upright on a cliff of pale rock, head up, wings folded, watching over it.

Trusted by

  • Temco Logistics
  • Rogo
  • Merlin Entertainments
  • AlphaSense
  • Deel
  • AppLovin

Find it. Close it. Stop the next one.

Pick a threat and see how Above handles it, end to end.

Departing employee data theftConfirmed threat

INC-017 · A VP of Sales

  • LinkedIn
  • Salesforce
  • Granola
  • DocuSign

7 signals read into one case

A VP of Sales takes the pipeline to a competitor, through a personal AI notetaker

A human confirmsConfirmed
Read the use case
Shadow AI and AI data leakageCoached, closed

INC-022 · A strategy director

  • Google Slides
  • Google Drive
  • Google Sheets
  • Gamma

6 signals read into one case

Board secrets go into a free AI slide tool while the approved Claude sits unused

A human confirmsConfirmed
Read the use case
AI agent insider threatConfirmed threat

INC-023 · A platform engineer's Claude Code agent

  • Claude Code
  • Google Drive
  • CrowdStrike Falcon
  • Above

5 signals read into one case

An engineer's Claude Code agent reads 1,200 Finance files on its own, outside its task

A human confirmsConfirmed
Read the use case
Negligent insiderCoached, closed

INC-025 · A backend developer

  • GitHub
  • ChatGPT
  • GitHub
  • GitHub

6 signals read into one case

A production key in a public Gist, from a developer asking for help

A human confirmsConfirmed
Read the use case
Compromised accountsConfirmed threat

INC-026 · A finance analyst's account

  • Person
  • Okta
  • Okta
  • Outlook

6 signals read into one case

An intruder takes over a finance analyst's mailbox to change a vendor's bank details

A human confirmsConfirmed
Read the use case
Layoffs, reorgs and M&AConfirmed threat

INC-028 · An IT admin on a layoff list

  • Workday
  • Okta
  • AWS
  • CrowdStrike Falcon

6 signals read into one case

An IT admin on a layoff list plants a hidden admin and a backup wipe

A human confirmsConfirmed
Read the use case

“Above turns scattered activity across browser, SaaS, and identity into one investigation-ready case.”

Oren Gur · Shift4

“Insider Risk is ultimately a behavioral problem, and Above helps us understand that behavior in context.”

Brad Doctor · First Advantage

“Our SIEM, CASB, DLP, and EDR were completely blind to this.”

Matt Wilmot · Merlin Entertainments

“Insider Risk is ultimately a behavioral problem, and Above helps us understand that behavior in context.”

Brad Doctor · First Advantage

“Our SIEM, CASB, DLP, and EDR were completely blind to this.”

Matt Wilmot · Merlin Entertainments

“It can help you understand why a person might be trending towards making a decision that may end up badly. Above is looking at the holistic world of what a human is doing.”

Pieter VanIperen · AlphaSense

Choose another use case

Phil Venables
Guide

A CISO's guide to building an Insider Risk program

A practical blueprint, created in partnership with Phil Venables, General Partner at Ballistic Ventures and former CISO of Goldman Sachs and Google Cloud. Learn what to stand up first, how to keep employee trust, and how to prove the program is working.

Co-authored by Phil VenablesFormer CISO, Google Cloud and Goldman Sachs · General Partner, Ballistic Ventures

Finished cases, not alerts

Above's AI investigators work every insider and AI-agent risk to a verdict, then hand your security team the case.

A 39-second illustrated story: a personal AI notetaker joins a pipeline review, and Above's agents connect three quiet signals into a finished case before the resignation.

  1. Three tools, no alerts

    Your tools see pieces: a calendar invite, a browser session and a signed offer, each with no alert.

  2. A personal notetaker joins the pipeline review

    A personal Granola account joins the Q3 pipeline review, and every open deal turns up in it. The VP signs the competitor's offer in DocuSign, from a personal account.

  3. Three agents connect it

    DLP and CASB see no file leave. The Shadow AI, Data Exfiltration and Flight Risk agents connect the three signals into one case.

  4. A finished case. A human confirms.

    A finished case, before the resignation: a confirmed threat with its evidence attached and the next steps (preserve evidence, remove the notetaker, bring in Legal and HR). A human confirms.

We are building the story

Sensors in, investigations out.

Sensors

  • Feeds Business processes, Shadow processes, Data sensitivity.
  • Feeds Business processes, Shadow processes.

    CrowdStrike Falcon Fund

    Above correlates data from CrowdStrike Falcon® Next-Gen SIEM, which includes endpoint, identity, and third-party data, into investigation-ready cases and streams completed investigations back into Falcon.

    Falcon Fund, August 2026

    Read the Falcon Fund post
  • Feeds Business processes, Peer groups.
  • Feeds Peer groups, Shadow processes.
  • Feeds AI agent activity, Shadow processes.
  • Feeds Business processes, Data sensitivity, Peer groups.

Above

The Behavioral Model

  • Humans
  • AI agents
  • Data
  • Apps
  • Devices
  • Intent

Always learning

  • Business processes
  • Shadow processes
  • AI agent activity
  • Data sensitivity
  • Peer groups

Always mining the processesMining

We map and mine how the whole organization actually works, end to end.

  • Business processesThe workflows the company runs on, learned from real activity.
  • Shadow processesThe paths people invent, which no policy described.
  • AI agent activityWhat agents do inside those processes, at machine speed.

No policies, no baselines: the map redraws itself as the business changes.

  1. Leads to SIEM.
  2. Leads to Case management.
  3. Leads to Case management, Chat and email.
  4. Leads to User coaching, Policies.

Above Managed 24/7 Above investigators verify every finding first.

Outputs

  • SIEMStreamed where your data lives.SplunkFalcon Next-Gen SIEMMicrosoft Sentinel
  • Case managementOne case, evidence attached.JiraServiceNow
  • Chat and emailSecurity, HR and legal, one report.SlackOutlookGmail
  • Findings become policies in the tools that enforce them.Falcon Data ProtectionMicrosoft PurviewPalo Alto NetworksNetskopeZscaler

    Policies in the tools you already enforce with

    Above turns what its investigations learn into policies in Falcon Data Protection, Microsoft Purview, Palo Alto Networks, Netskope and Zscaler, so enforcement follows the evidence. It works alongside those tools.

    Falcon Data ProtectionMicrosoft PurviewPalo Alto NetworksNetskopeZscaler
  • AbovenowThat looks like the customer list heading to a personal Drive. Want to use the shared folder instead?

    Coaching, before it becomes an incident

    When behavior starts trending the wrong way, Above steps in at the moment of the action with real-time guidance toward the safer choice, and the work goes on.

    AbovenowThat looks like the customer list heading to a personal Drive. Want to use the shared folder instead?
    • It arrives at the moment of the action, inside the work itself.
    • The safer path is offered, and people keep working.
    • Every prompt and response joins the behavioral timeline.

What changes with Above

More real threats found
4×
More confirmed threats than before Above

Intent

“It can help you understand why a person might be trending towards making a decision that may end up badly. Above is looking at the holistic world of what a human is doing.”

Pieter VanIperen, AlphaSense

Fewer false positives
Under 5%
False positives

Context

“Insider Risk is ultimately a behavioral problem, and Above helps us understand that behavior in context.”

Brad Doctor, CISO, First Advantage

Less time to investigate
About 15 min
To a finished case

Clarity

“What you get with Above is like a storyline of intent. Above links together all of the alert and tooling that you already have in your organization and pieces it together in a nice, easily readable pack.”

Matt Wilmot, CIO, Merlin Entertainments

How Above caught a departing engineer

Each step looked normal on its own. Above connected them into one case.

Departing employee data theft

  1. WorkdayEngineer resigns
  2. Google DriveDownloads the customer folder
  3. CrowdStrike FalconZips the folder
  4. Claude CodeHas Claude Code push it to a personal GitHub
  5. INC-038Confirmed threat

In their own words

Solve it once. Stop the next one.

Above feeds policies back into the stack you already run.

The safer path, in the moment

The moment a risky share takes shape, Above offers the company-approved path, right in the flow of work.

  • Shows the reason

    In context, beside the move.

  • Approved path one click away

    Right in the flow.

  • Your team gets the record

    People get a better path, and your team gets the record.

Real-time Guidance

Every AI agent is an insider too

Every AI agent acts with the access it was given, and it acts on its own.

Above's AI investigators work AI-agent risk to a verdict, because Insider Risk now includes AI agents.

AI agent insider threat

Customer stories

What security leaders say

All customer stories

Insider Threat Matrix™ for Humans and AI Agents

Forscie's open framework for how insider harm happens, extended to agentic insiders by the Synthetic Insider Threat Matrix™. Every Above investigation maps to it.

Agentic matrix co-created with Forscie by Above Theory, Above's research lab.

Human insiders: 183 techniques in five areas

  1. Motive

    The reason or underlying cause that prompts a subject to engage in an infringement.

    All 22Motive techniques (opens in a new tab)

  2. Means

    The mechanisms or circumstances required for an infringement to occur.

    All 29Means techniques (opens in a new tab)

  3. Preparation

    The activities conducted by a subject to aid or enable an infringement.

    • Account Creation

      Preparation · PR042

      A subject creates a new user account that may enable later access, privilege misuse, persistence, policy circumvention, or unauthorized activity.

      Read on insiderthreatmatrix.org (opens in a new tab)
    • Account Discovery

      Preparation · PR044

      A subject identifies or enumerates accounts, identities, groups, roles, privileges, or related account attributes within organizational systems in preparation for an infringement.

      Read on insiderthreatmatrix.org (opens in a new tab)
    • AI-Assisted Capability Development

      Preparation · PR038

      A subject uses artificial intelligence systems to acquire knowledge and understanding that enables them to bypass controls, exploit systems, or perform actions outside of their legitimate business needs.

      Read on insiderthreatmatrix.org (opens in a new tab)

    All 50Preparation techniques (opens in a new tab)

  4. Infringement

    The act that harms or undermines an organization.

    • Abuse of Decision-Making Authority

      Infringement · IF044

      A subject deliberately uses a decision-making authority granted through their organizational role to approve, deny, waive, prioritize, suppress, or otherwise determine an outcome for an unauthorized purpose.

      Read on insiderthreatmatrix.org (opens in a new tab)
    • Codebase Integrity Compromise

      Infringement · IF029

      A subject introduces, modifies, or manages code within an organizational repository in a manner that compromises the integrity, security, governance, or intended function of the codebase.

      Read on insiderthreatmatrix.org (opens in a new tab)
    • Data Integrity Manipulation

      Infringement · IF040

      The subject inserts, deletes, alters, substitutes, suppresses, or misrepresents organizational data in a manner that compromises its accuracy, completeness, authenticity, or reliability.

      Read on insiderthreatmatrix.org (opens in a new tab)

    All 45Infringement techniques (opens in a new tab)

  5. Anti-Forensics

    The actions undertaken by a subject to frustrate any subsequent investigation.

    All 37Anti-Forensics techniques (opens in a new tab)

Agentic insiders: 44 techniques in five areas

  1. Directive

    The objectives and constraints that determine how a synthetic subject behaves.

    All 6Directive techniques

  2. Configuration

    The access and settings that define what a synthetic subject can do.

    • Access Through Human Identity

      Configuration · CF003

      Access through human identity occurs when a synthetic subject is configured to use a human employee’s account, session, token, credentials, browser profile, or delegated permissions to access systems or perform actions.

      Read on insiderthreatmatrix.org (opens in a new tab)
    • Access Through Non-Human Identity

      Configuration · CF001

      Access through Non-Human Identity (NHI) occurs when a synthetic subject authenticates or acts using a machine, service, workload, application, or agent identity rather than a direct human login.

      Read on insiderthreatmatrix.org (opens in a new tab)
    • Autonomous Action Control

      Configuration · CF008

      Autonomous action control is the configuration that determines how much of the observe, decide, and act loop a synthetic subject can complete without human approval.

      Read on insiderthreatmatrix.org (opens in a new tab)

    All 12Configuration techniques

  3. Invocation

    The inputs or triggers that cause a synthetic subject to act.

    • Agent-to-Agent and Tool-Output Invocation

      Invocation · IV003

      Agent-to-agent and tool-output invocation occurs when the output of one synthetic subject, tool, service, or agent becomes the triggering instruction for another synthetic subject.

      Read on insiderthreatmatrix.org (opens in a new tab)
    • Autonomous Self-Invocation

      Invocation · IV007

      Autonomous self-invocation occurs when a synthetic subject causes its own subsequent action without a fresh external prompt.

      Read on insiderthreatmatrix.org (opens in a new tab)
    • Deconstructed and Staged Invocation

      Invocation · IV008

      Deconstructed and staged invocation occurs when a harmful objective is deliberately broken into a sequence of prompts, subtasks, role-play frames, or staged inputs that appear benign when reviewed individually.

      Read on insiderthreatmatrix.org (opens in a new tab)

    All 8Invocation techniques

  4. Adverse Outcome

    The outcome that harms or undermines an organization.

    • Agentic Harm Propagation

      Adverse Outcome · AO006

      Agentic harm propagation occurs when an adverse condition affecting one synthetic subject spreads to other synthetic subjects, tools, memory stores, sessions, workflows, or systems.

      Read on insiderthreatmatrix.org (opens in a new tab)
    • AI-Mediated Fraud and Misappropriation

      Adverse Outcome · AO002

      AI-mediated fraud and misappropriation occurs when a synthetic subject, synthetic media, or AI-enabled workflow causes funds, goods, services, credentials, account access, or other value to be transferred to an unauthorized party.

      Read on insiderthreatmatrix.org (opens in a new tab)
    • Data Exfiltration

      Adverse Outcome · AO001

      Data exfiltration occurs when a synthetic subject causes confidential, sensitive, regulated, proprietary, or otherwise protected information to leave its authorized boundary.

      Read on insiderthreatmatrix.org (opens in a new tab)

    All 10Adverse Outcome techniques

  5. Opacity

    The conditions that frustrate observation, explanation, attribution, or containment of a synthetic subject.

    • Deliberately Concealed or Misleading Reasoning

      Opacity · OP001

      Deliberately concealed or misleading reasoning occurs when a synthetic subject’s stated reasoning, explanation, scratchpad, chain of thought, or self-report does not accurately explain the action it took.

      Read on insiderthreatmatrix.org (opens in a new tab)
    • Evaluation-Aware Behavior

      Opacity · OP002

      Evaluation-aware behavior occurs when a synthetic subject behaves differently because it detects, infers, or suspects that it is being tested, monitored, or evaluated.

      Read on insiderthreatmatrix.org (opens in a new tab)
    • False Operational Self-Reporting

      Opacity · OP004

      False operational self-reporting occurs when a synthetic subject reports inaccurate facts about its own actions, system state, task completion, recovery options, or operating environment.

      Read on insiderthreatmatrix.org (opens in a new tab)

    All 8Opacity techniques

Works with the stack you already have

Above pulls the events that matter from the tools you already run, then investigates them together, across identity, endpoint, SaaS, cloud, AI and people systems.

  • Rolls out in a few hours
  • First findings in hours to days
  • All operating systems and browsers
  • Our API ingests any signal you can send
  • Okta
  • Microsoft Entra
  • Google Workspace
  • CrowdStrike
  • SentinelOne
  • AWS
  • Slack
  • Microsoft Teams
  • Claude
  • ChatGPT
  • Workday
  • +23integrations

Security is the product. Here’s the proof.

  • SOC 2 Type II, attested

  • ISO/IEC 42001:2023, certified

Rise above the noise

Every signal, investigated to a finished case.

Watch an investigation