Insider Risk vs. insider threat
Insider Risk is the broad set of ways people, contractors and AI agents with authorized access can harm an organization, accidentally or deliberately. An insider threat is the narrower case: a person or entity with the intent to do harm, whether malicious, coerced or planted.

On this page6 sections
The difference at a glance
Phil Venables, former CISO of Goldman Sachs and Google Cloud, separates the two terms in Redefining insider threat: “Insider threat is the person or entity with the intent to do harm: malicious, coerced, or deliberately planted. Insider Risk is the broader universe of ways insiders can create harm for the organization, accidental or deliberate.”
| Aspect | Insider Risk | Insider threat |
|---|---|---|
| What it is | Every way trusted access can cause harm | A person or entity with the intent to do harm |
| Intent | Not required: much of it carries no malicious intent | Required: malicious, coerced or deliberately planted |
| Typical examples | Data moved into an unsanctioned tool to meet a deadline, a file sent to the wrong person | Data taken by a departing employee, a bribed contractor, a fake IT worker |
| Who can be involved | Employees, former employees, contractors, vendors, partners, service accounts, OAuth apps, AI agents | The same people and identities; what differs is intent |
| What a program does next | Contain it, fix the broken guardrail, coach the person | Investigate without tipping them off, preserve evidence, bring in HR and legal early |
Every insider threat is Insider Risk, but most Insider Risk is not a threat. Venables’s point is that a program built only to find explicit threats “ignores the much larger class of events where most incidents actually originate.”
Who counts as an insider
CISA’s Insider Threat Mitigation Guide (opens in a new tab) defines an insider as “any person who has or had authorized access to or knowledge of an organization’s resources, including personnel, facilities, information, equipment, networks, and systems.” CISA and NIST both define the threat as an insider using that authorized access, intentionally or unintentionally (NIST: “wittingly or unwittingly”), to do harm.
Israel Bryski, writing from his years as a CISO in What “insider” actually means in 2026, argues the definition now has to reach past the HR roster. His working draft: “An insider is any human or non-human identity, employee, contractor, vendor, service account, OAuth-connected application, or autonomous agent that holds, or has recently held, authorized access to the organization’s systems, data, or decision-making processes, and whose action or inaction could result in harm to the organization, intentionally or unintentionally.”
Types of insider threat
Insider Threat 101 groups insiders by intent and by the data they can reach:
- Malicious insiders act to harm the organization: stealing trade secrets, intellectual property or customer data, or sabotaging systems, often for money, revenge or ideology.
- Negligent insiders mean no harm but knowingly ignore policy for speed or convenience, such as sharing passwords or putting confidential data into a generative AI prompt.
- Unintentional insiders make mistakes without realizing it: falling for a phishing attack, emailing information to the wrong recipient, misconfiguring an app.
- Third-party insiders are contractors, vendors and partners with authorized access, who can be malicious or accidental.
- Collusive threats are insiders working with others, inside or outside, such as criminals or nation-state actors paying for credentials or data.
By the stricter definition above, only the malicious and collusive cases are insider threats; negligent and unintentional insiders are the wider Insider Risk. Venables adds three personas the classic model misses: the third-party insider who abuses access that was granted on purpose, the instant insider whom extortion or bribery turns overnight (“Trust is not a control against blackmail”), and the fake insider who joins only to gain access, such as fake IT workers.
Insider Risk without a threat
Filip Stojkovski, after 13 years building Insider Risk programs, names the unintentional risks he sees most in Unintentional risk shows you what’s broken: shadow IT, data leaks from people trying to get the job done, and shadow AI. “Pasting a customer contract into a random chatbot to ‘make it shorter’ is the new forwarding to personal Gmail.” None of these involve intent to harm, and all of them expose data.
How programs address each
For unintentional risk, Stojkovski’s loop is to contain the exposure, ask why the person could do it and find the broken guardrail, fix it with the team that owns the control, and notify or coach the person. These cases usually tell you something in your controls is not working as expected.
Intentional threats are harder, because the insider knows how the organization works and may know its detections. Some actions trigger default DLP rules, some show up in EDR or SIEM alerts, canary tokens catch people looking where they shouldn’t, and for everything else you need an Insider Risk platform that monitors activity and employee behavior. The investigation has to run without alerting the person, and HR and legal should come in early so the evidence holds up. Turning SIEM and DLP alerts into Insider Risk cases walks through how.
Between the two sits the messy middle: the same upload to a personal drive can be either, and a compromised account looks exactly like an insider until you find the attacker behind it. Intent is what the investigation tells you, not the label you start with. That is the problem Above is built for: Above’s AI investigators weigh role, access, timing, history and what changed to tell harmless activity from misuse, negligence or malice, and real-time guidance coaches people at the moment of risk instead of alerting after.
Where AI agents fit
AI agents have valid identities, real permissions and access to company data, “pretty much the definition of an insider,” in Stojkovski’s words. Bryski describes an agent with a service account that could read deal memos that people in adjacent business units could not, with nobody in security owning its behavior.
Agents change the intent question. “An agent has no intent, they just have directives.” In the SITM’s own example, which Stojkovski retells, an employee asks an internal assistant to summarize a vendor proposal, hidden text in the file redirects the assistant, and the data leaves over approved email under the employee’s identity. The employee did nothing wrong, and the audit log blames them anyway.
The Synthetic Insider Threat Matrix, built by Above Theory with Forscie, extends the Insider Threat Matrix™ to AI agents and breaks an incident into five layers: the directive (what the agent is told to do), its configuration (what it can reach), the invocation (what triggered it), the adverse outcome, and the opacity that makes it hard to see or prove afterward. For how Above covers agents, see agentic AI.
Further reading
Further readingRedefining insider threatPhil Venables on redefining insider threat for the agentic era: how people, contractors and AI agents now work, and how outside threats coerce trusted insiders.
Further readingWhat "insider" actually means in 2026A former CISO on why the textbook 'insider' definition no longer holds in 2026, and how AI agents and modern access are reshaping Insider Risk.
Further readingUnintentional risk shows you what's brokenAfter 13 years building Insider Risk programs: unintentional risk shows you what is broken, and intentional risk tests whether you can see it at all.
Questions teams ask
What is the difference between Insider Risk and insider threat?
Insider Risk is every way someone with authorized access can harm an organization; an insider threat is the person or entity with the intent to do harm. Most Insider Risk carries no malicious intent, such as data moved into an unsanctioned tool to meet a deadline. Every insider threat is Insider Risk, but not the reverse.
What is an insider threat?
An insider threat is a person or entity with authorized access and the intent to harm the organization: malicious, coerced, or deliberately planted. Examples include an employee stealing trade secrets before leaving, a contractor bribed by an outside group, and a fake IT worker hired to gain access. The threat uses legitimate access the organization itself granted.
What are the types of insider threats?
Insider threats are usually grouped by intent into five types: malicious insiders who mean harm, negligent insiders who ignore policy for convenience, unintentional insiders who make mistakes, third parties such as contractors and vendors, and collusive threats where insiders work with others. Only malicious and collusive insiders intend harm; the rest are wider Insider Risk.
Is an accidental data leak an insider threat?
An accidental data leak is Insider Risk rather than an insider threat, because nobody intended harm. Emailing a file to the wrong recipient or pasting a contract into an unapproved chatbot still exposes data. Programs handle it by containing the exposure, finding the broken guardrail that allowed it, fixing it, and coaching the person involved.
Can an AI agent be an insider?
An AI agent is an insider when it holds a valid identity, real permissions and access to company data, just like an employee. Agents have no intent of their own, only directives, so an agent can leak data under an employee's identity when a hidden instruction redirects it. The Synthetic Insider Threat Matrix™ maps how this happens.
How does an Insider Risk program handle intentional and unintentional cases?
An Insider Risk program handles unintentional cases by containing the exposure, fixing the guardrail and coaching the person, and intentional cases by investigating quietly, preserving evidence and bringing in HR and legal early. The same action can be either, so intent is what the investigation establishes, not the label a case starts with.
Ready to make your move?
Most insider threats are preventable. With Above, every signal becomes a finished case before it reaches your team.

