Skip to content

Human Error and Insider Threat Statistics 2026: Mistakes, Misuse and Malicious Insiders

Human error and insider threat statistics for 2026 from Verizon DBIR, IBM, Ponemon/DTEX, KnowBe4, Proofpoint, Mimecast, UK DSIT and Cyberhaven.

Azad KhanBy , Senior GRC & Compliance16 min read

Penetration Testing ServicesExternal network from $4,200, web app from $5,200Fixed scope, free retest, no hourly billing.

Get my fixed price

This page collects the measured numbers on how people inside an organization cause breaches: errors, social engineering, training, negligent and malicious insiders, insider costs and containment time, privilege misuse, and company data pasted into unapproved AI tools. The figures come from Verizon’s Data Breach Investigations Report, IBM’s Cost of a Data Breach Report, the Ponemon Institute and DTEX Cost of Insider Risks study, KnowBe4, Proofpoint, Mimecast, the UK government’s Cyber Security Breaches Survey, Cyberhaven, Harmonic Security and a peer-reviewed field study of phishing training. Each figure links to its source.

How to cite: link to this page or to the primary source beside each figure. Survey figures carry their sample size.

Key numbers at a glance

1. The human element in breaches

Verizon counts a breach as involving the human element when a person had to be involved for it to progress: someone clicked an email, visited a website, picked up the phone or made an error. Its 2026 report covers more than 31,000 incidents and 22,000 confirmed breaches in 145 countries, from October 2024 to November 2025.

  • 62% of breaches involved the human element in the 2026 dataset, up from 60% in 2025. (Verizon 2026 DBIR Executive Summary)
  • The 2025 report put the figure at 60% against 61% a year earlier, which Verizon treats as the same within its 95% confidence tolerance. (Verizon 2025 DBIR)
  • Human error caused 26% of breaches and IT failure 23%; malicious or criminal attacks caused the other 51%. (IBM Cost of a Data Breach Report 2025, 600 organizations breached between March 2024 and February 2025)
  • 79% of CISOs name human risk as their organization’s biggest cyber vulnerability, up from 66% in 2025. (Proofpoint 2026 Voice of the CISO, more than 1,600 CISOs at organizations with 1,000 or more employees)
  • Respondents estimate that a single insider-driven data exposure, loss, leak or theft would cost their organization $13.1 million on average. This is a survey estimate, not a measured loss. (Mimecast State of Human Risk 2026, 2,500 IT and IT security decision makers in nine countries)

Human element and internal actors by industry, 2026

Sector Human element in breaches Internal actors in breaches
Public administration 69% 44%
Educational services 68% 22%
Financial and insurance 65% 12%
Retail 58% 1%
Manufacturing 56% 5%
Healthcare 54% 19%
Small and medium-sized businesses 45% not listed (external 100%)

Source: Verizon 2026 DBIR Executive Summary, industry highlights.

“Human element” and “insider” are not the same thing. Most human-element breaches start with an outsider who tricks an employee; internal actors cause the breach themselves, by mistake or on purpose.

2. Misdelivery, misconfiguration and other errors

  • Verizon recorded 1,476 incidents and 1,449 confirmed breaches in its Miscellaneous Errors pattern. 98% were caused by internal actors and 95% exposed personal data. The three most common errors were misdelivery, misconfiguration and publishing errors. (Verizon 2025 DBIR)
  • Among 2,199 breaches caused by internal actors, unintentional errors outnumbered deliberate privilege misuse by roughly 2 to 1. End users made up 9% of all actor varieties, and 72% of the actions involving end users were misdelivery. (Verizon 2025 DBIR)
  • In public administration, internal actors were behind 44% of breaches in 2026, driven by misdelivery “due to the sheer volume of correspondence.” In healthcare, Verizon calls staff mistakes and misconfigurations a chronic source of breaches. (Verizon 2026 DBIR Executive Summary)
  • A breach caused by insider error cost $3.62 million on average and took 213 days to identify and contain (162 to identify, 51 to contain). (IBM 2025)
  • At third-party cloud providers, weak passwords and permission misconfigurations took almost eight months to resolve half of all findings. (Verizon 2026 DBIR Executive Summary)

Errors are the insider category that testing finds directly: an overshared SharePoint site or a mail rule forwarding to a personal address is a finding, not a behavior. A Microsoft 365 security assessment looks for exactly these sharing and forwarding settings.

3. Social engineering susceptibility

  • Social engineering was the third most common breach pattern in 2026, at 16% of breaches. Phishing was present in 16% of all breaches, unchanged from the previous year, and pretexting rose to 6%. (Verizon 2026 DBIR Executive Summary)
  • In phishing simulations, the median click rate for mobile-centric lures such as voice and text messages was 40% higher than for email. (Verizon 2026 DBIR Executive Summary)
  • Phishing was the most common initial attack vector at 16% of breaches, at an average cost of $4.8 million. (IBM 2025)
  • 33.2% of employees click or act on a simulated phishing email before any training. Baselines were highest in healthcare and pharmaceuticals (42.7%), insurance (38.1%) and retail and wholesale (36%); North America started at 34.5%. Organizations with 10,000 or more employees started at 39.5%, small businesses at 24.7%. (KnowBe4 2026, 42 million simulations, 14.8 million users, 64,000 organizations)
  • In an eight-month field trial at a large healthcare organization, 56% of users (11,077 of 19,789) clicked at least one simulated phishing link, against 9.7% in the first month. Some lures drew clicks from 1 to 2% of users, others from upwards of 30%. (Ho, Mirian, Luo et al., IEEE Symposium on Security and Privacy 2025)
  • 71% of working adults admitted taking a risky action such as reusing a password, and 96% of them knew it was risky. Their reasons: convenience (44%), saving time (39%) and urgency (24%). (Proofpoint 2024 State of the Phish, 7,500 employees and 1,050 security professionals in 15 countries)
  • Phishing was the most common breach or attack for UK businesses (38%) and the most disruptive for 69% of those attacked; 12% experienced impersonation. (UK DSIT Cyber Security Breaches Survey 2025/2026, 2,112 businesses)

One month’s click rate says little about who will click over a year, and the lure matters as much as the audience. Our phishing attack statistics cover attack volume and BEC losses; spear phishing explains the researched form aimed at one named person.

See what a real pentest finds. Read our sample report, then get a fixed quote in writing.

4. Does security awareness training work?

Vendor simulation data shows large drops in clicking; a randomized trial and Verizon’s data show small effects on clicking and a large effect on reporting.

  • After 90 days of training the simulated phish click rate fell to 20.1%, and after a year to 4.2%. KnowBe4 states the reductions as 40% and 79%. (KnowBe4 2026, 14.8 million users)
  • Employees trained within the previous 30 days reported phishing at about 21%, against a base rate of 5%, a fourfold increase. The effect of recent training on click rate was much smaller, about 5% relative per training. In the median organization about 1.5% of employees keep clicking regardless. (Verizon 2025 DBIR, data from more than 7,000 organizations)
  • The randomized trial found no significant relationship between recently completing annual awareness training and failing a phishing simulation. Users given embedded training had a failure rate 1.7% lower on average (an absolute difference), and in the worst case 37% to 51% of training sessions had no engagement: users closed the page immediately. (Ho, Mirian, Luo et al., IEEE S&P 2025, more than 19,500 employees)
  • Employee training reduced the average breach cost by $192,266. (IBM 2025)
  • Only 19% of UK businesses gave staff any cyber security training in the past year (54% of medium and 84% of large businesses), and 22% tested staff with exercises such as mock phishing. After a breach, 31% made people or training changes. (UK DSIT 2025/2026, 2,112 businesses)
  • 53% of organizations run regular awareness training and 52% continuously monitor for policy violations, but only 28% do both. (Mimecast 2026, 2,500 respondents)

What this means: train to raise the report rate, and build controls that assume the click. A phishing test against your own domain gives a measured baseline for click, credential-entry and report rates.

5. Negligent, malicious and compromised insiders

CISA defines an insider as anyone who has or had authorized access to or knowledge of an organization’s resources. It separates unintentional threats (negligence and accidents) from intentional, collusive and third-party threats. (CISA, Defining Insider Threats) Ponemon’s categories map onto the same split.

  • Organizations averaged 25.4 insider incidents in 2025. (DTEX 2026)
  • 53% of incidents were due to employee negligence, 27% to malicious or criminal insiders and 20% to credential theft. (Ponemon 2026 Cost of Insider Risks, 8,750 IT and security practitioners at 354 organizations)
  • 68% of organizations had between 21 and more than 40 insider incidents in a year, up from 57% in the 2024 study. (Ponemon 2026)
  • Among organizations that suffered material data loss, CISOs blamed malicious or criminal insiders (46%), careless insiders (38%) and compromised insiders (38%), and 93% said departing employees played a role. The share reporting material data loss fell from 66% to 53%. (Proofpoint 2026 Voice of the CISO, more than 1,600 CISOs)
  • 42% of organizations saw more threats from malicious insiders in the past year, the same 42% that saw more negligent employee incidents. (Mimecast 2026, 2,500 respondents)
  • Mimecast states that 8% of employees account for 80% of security incidents, a figure it sources to its own whitepaper, The Size and Shape of Workforce Risk. (Mimecast 2026)
  • 17% of organizations reported no insider attacks in 2024, down from 40% in 2023; 48% said insider attacks became more frequent. (Cybersecurity Insiders 2024 Insider Threat Report, 413 IT and security professionals, sponsored by Gurucul)
  • 93% of security leaders find insider threats as hard or harder to detect than external attacks, and only 23% strongly believe they can detect and prevent them proactively. (Cybersecurity Insiders 2025 Insider Risk Report, 635 respondents)

Insider incident types compared, 2025

Type Share of incidents Incidents per organization Cost per incident Annual cost
Negligent or mistaken 53% 13.8 $747,107 $10.3 million
Malicious or criminal 27% 6.3 $742,125 $4.7 million
Credential theft (compromised) 20% 5.3 $842,462 $4.5 million

Source: Ponemon 2026 Cost of Insider Risks. Negligence costs rose 17% year over year; credential theft rose from 4.8 incidents and $779,707 per incident in the 2024 study.

6. Insider incident costs and containment time

  • The average annual cost of insider risk reached $19.5 million in 2025, up from $17.4 million and 20% over two years. (DTEX 2026, 354 organizations with 500 to more than 75,000 employees)
  • North American companies averaged $24 million, European companies $18.6 million. Health and pharma averaged $28.8 million and technology and software $24.2 million. Organizations with more than 75,000 employees spent $28.4 million; those with fewer than 500, $8.9 million. (Ponemon 2026)
  • Containing an incident took 67 days on average in 2025, down from 81 days in 2024 and 86 in 2023. Only 13% of incidents were contained in less than 30 days. (Ponemon 2026; DTEX 2026)
  • Organizations that contained incidents within 30 days averaged $14.2 million a year; those taking more than 90 days averaged $21.9 million. (Ponemon 2026)
  • Containment was the most expensive activity at $247,587 per incident; escalation was the cheapest at $39,728. (Ponemon 2026)
  • A breach that started with a malicious insider cost $4.92 million and took 260 days to identify and contain (194 and 66), the second longest of any vector after supply chain compromise. (IBM 2025)
  • 63% of organizations run a dedicated insider risk program; those programs prevented an average of seven incidents and avoided $8.2 million in costs. Insider risk spending grew from 8.2% of IT security budgets in 2023 to about 19% in 2025. 64% raised insider risk budgets in 2025, yet 45% still call funding insufficient. (DTEX 2026)

Time is the cost driver: by simple arithmetic, the gap between fast and slow containment in Ponemon’s data is $7.7 million a year. An internal network penetration test starts from a normal employee account and shows how far it reaches and whether anyone notices.

7. Privilege misuse

  • Verizon recorded 825 privilege misuse incidents and 757 confirmed breaches. 90% were caused by internal actors and 10% by partners. Motives were financial in 89% of breaches, espionage in 10% and a grudge in 5%; personal data was taken in 72%. (Verizon 2025 DBIR)
  • Verizon saw a convergence between rank-and-file end users and system administrators as the insiders behind misuse, a shift it calls concerning given the access administrators hold. (Verizon 2025 DBIR)
  • In public administration, system intrusion, miscellaneous errors and privilege misuse together made up 80% of breaches in 2026. (Verizon 2026 DBIR Executive Summary)

Most misuse cases are employees copying data they were already allowed to see. The control is access scoped to the job and reviewed when the job changes, and the test is whether a standard account can reach more than it should. Physical access is the other path an insider, or someone who tailgates one, can take: our physical security assessment checklist lists what to check, and physical penetration testing tests it.

8. Shadow IT and GenAI data leakage by employees

  • 45% of employees are now regular users of AI on their corporate devices, up from 15% a year earlier, and 67% of users access AI services with non-corporate accounts on corporate devices. Shadow AI became the third most common non-malicious insider action in Verizon’s data loss prevention dataset, a fourfold increase. (Verizon 2026 DBIR Executive Summary)
  • Source code was the most common data type sent to external GenAI models, and research and technical documentation appeared in 3.2% of DLP policy violations. (Verizon 2026 DBIR Executive Summary, 858,440 DLP events)
  • 39.7% of AI interactions involve sensitive data; the average employee enters sensitive data into an AI tool once every three days. Personal accounts carry 32.3% of ChatGPT use, 24.9% of Gemini, 58.2% of Claude and 60.9% of Perplexity. 82% of the 100 most-used GenAI apps are rated medium, high or critical risk. (Cyberhaven 2026, 222 companies)
  • Across 22,458,240 enterprise prompts in 2025, Harmonic detected 579,113 sensitive data exposures. 16.9% went through personal free-tier accounts. Code was 26.5% of exposures and legal content 22.3%; ChatGPT accounted for 71.2% of the risk. (Harmonic Security, 2026, 665 AI tools tracked)
  • The average enterprise saw 23 previously unknown GenAI tools come into use in a single quarter. (Harmonic Security, Q2 2025, 1 million prompts and 20,000 files)
  • Shadow AI was involved in 20% of breaches, and organizations with high levels of it paid $670,000 more per breach ($4.74 million against $4.07 million). 63% of organizations had no AI governance policy. (IBM 2025)
  • In IBM’s 2026 study, unapproved AI tools figured in 43% of security incidents, more than double the previous year’s share, as reported by Help Net Security. IBM’s public summary does not state the figure. (IBM Cost of a Data Breach Report 2026, 602 organizations)
  • 73% of organizations worry unauthorized AI use is creating invisible data exfiltration paths, and only 19% classify AI agents as equivalent to human insiders. (DTEX 2026)
  • 76% of CISOs believe employees are likely to use AI in ways that could expose sensitive data. (Proofpoint 2026 Voice of the CISO, more than 1,600 CISOs)

Shadow AI is the old shadow IT problem, except the tool keeps what it is given. Much of the exposure runs through personal accounts outside corporate logging, so offering an approved tool with a corporate login is part of the control.

What to expect in 2027

These are trends in the data above, labelled as such; only the budget item is the source’s own forecast.

  • Trend: shadow AI keeps growing. Regular AI use on corporate devices tripled in a year, from 15% to 45% (Verizon 2026 DBIR), and the share of incidents involving unapproved AI more than doubled in IBM’s 2026 study, as reported by Help Net Security.
  • Forecast: insider budgets rise. 70% of organizations expect insider risk budgets to rise again in 2026, and 28% expect increases of 10% or more. (DTEX 2026)
  • Trend: containment gets faster while costs rise. Average containment fell from 86 days (2023) to 81 (2024) to 67 (2025) across three studies, while annual cost rose to $19.5 million. (Ponemon 2026; DTEX 2026)
  • Trend: social engineering moves to the phone. Mobile lures draw a 40% higher median click rate than email, and pretexting is a growing route into ransomware (Verizon 2026 DBIR). Testing that covers only email will undercount the risk.
  • Trend: CISOs rank people first. The share naming human risk as the top vulnerability rose from 66% to 79% in a year. (Proofpoint 2026)

Part of our cybersecurity statistics hub. See also phishing attack statistics and data breach statistics.

Sources

Frequently asked questions

What percentage of data breaches are caused by human error? It depends on the definition. Verizon’s 2026 DBIR finds the human element, which includes being tricked, in 62% of breaches. Counting only mistakes, IBM’s 2025 report attributes 26% of breaches to human error, with another 23% to IT failure.

What percentage of insider threats are negligent rather than malicious? In Ponemon’s 2026 study of 354 organizations, 53% of insider incidents were caused by negligent employees, 27% by malicious insiders and 20% by credential theft. Verizon’s 2025 data agrees on the order: among breaches caused by internal actors, errors outnumbered deliberate misuse about 2 to 1.

How much does an insider threat cost? $19.5 million a year per organization on average in 2025 (Ponemon and DTEX). Per incident, negligence costs $747,107, a malicious insider $742,125 and credential theft $842,462. A single breach started by a malicious insider costs $4.92 million on average (IBM 2025).

Does security awareness training reduce phishing clicks? Vendor simulation data shows large drops (KnowBe4: 33.2% to 4.2% after a year). A randomized trial of more than 19,500 healthcare employees found a very small effect on clicking (an absolute difference of 1.7% on average), while Verizon found recent training raised the report rate from 5% to about 21%. The most consistent benefit is more reporting.

How often do employees put sensitive data into AI tools? Cyberhaven found sensitive data in 39.7% of AI interactions across 222 companies, about once every three days per employee. Harmonic found 579,113 sensitive exposures in 22.5 million prompts in 2025, 16.9% of them through personal free-tier accounts.

Breach causes and costs beyond insiders are in our data breach statistics.

Azad Khan

Written by

Azad Khan, Senior GRC & Compliance

26+ years in the field. Focus: GRC, ISO 27001 & PCI DSS consulting. Part of the team that scopes, tests, and reports for our clients. Full profile →

Find out what an attacker sees.

Tell us what to test and see your fixed price.

Prefer the full scoping questionnaire? 
Start the conversation

Get a quick proposal

We are online