Verizon published the 19th edition of its Data Breach Investigations Report on 19 May 2026. It is the largest yet: more than 31,000 real-world security incidents, of which more than 22,000 were confirmed data breaches, in organisations in 145 countries, occurring between 1 November 2024 and 31 October 2025. For the first time in the report's history, the most common way attackers got in was not a stolen password but a software vulnerability. Six months on, the figures still frame most of the security conversations we have with IT managers in Italy and Spain, so here is what the report actually says, with every number taken from the 2026 DBIR itself, and what a European mid-sized company should do about it. For the year's wider context, see our cybersecurity trends for 2026.
Key Takeaways
- Exploitation of vulnerabilities is now the top initial access vector: 31% of breaches, up from 20%. In Europe, the Middle East and Africa it reaches 47%.
- Patching got slower, not faster. Only 26% of KEV vulnerabilities were fully remediated (38% the year before), with a median of 43 days to full resolution (32 days before).
- Third parties are involved in 48% of breaches, up from 30%, and in 54% in EMEA.
- Ransomware is present in 48% of breaches, but 69% of victims did not pay, and about 96% of ransomware victims of known size were organisations with fewer than 1,000 employees.
- The human element is still there in 62% of breaches, and attacks are moving to phone and text messages.
The headline shifts in one table
| Metric (2026 DBIR) | 2026 | Previous edition |
|---|---|---|
| Exploitation of vulnerabilities as initial access vector | 31% | 20% |
| Credential abuse as initial access vector | 13% | 22% |
| Breaches with third-party involvement | 48% | 30% |
| Breaches involving ransomware | 48% | 44% |
| Human element present | 62% | 60% |
| KEV vulnerabilities fully remediated | 26% | 38% |
| Median days to fully remediate KEV vulnerabilities | 43 | 32 |
Two caveats from the report itself. First, the drop in credential abuse is partly a change of method: this year Verizon tracks pretexting as a separate initial access vector, and without that change credential abuse would have been 16%. Counted at any point in the attack rather than only at the start, credential abuse still leads, at 39% of breaches. Second, the initial access figures refer to breaches that were not errors or misuse, with a known entry point.
Exploitation first: what it means for patching
The rise from 20% to 31% is a 55% increase in one year. It would be less worrying if organisations were patching faster. They are not. In a study covering more than 13,000 organisations, Verizon looked at vulnerabilities listed in the CISA Known Exploited Vulnerabilities (KEV) catalog, the shortlist of flaws known to be used in real attacks:
- only 26% had been fully remediated, against 38% the year before;
- the share left entirely unremediated rose from 12% to 16%;
- the median time to full remediation grew from 32 to 43 days;
- the median organisation had 16 KEV vulnerabilities to patch in 2025, against 11 in 2024, almost 50% more.
The arithmetic is simple and uncomfortable. When a flaw in a VPN, firewall or file-transfer appliance is exploited within days of disclosure, a 43-day median means most organisations spend weeks exposed. The answer is not to patch everything faster, which nobody can do, but to patch the right things first: known-exploited vulnerabilities on systems reachable from the internet. Our article on exposed management interfaces shows what that looked like in practice in 2026, and the vulnerability management lifecycle explains how to build the process.
Europe is not the global average
The DBIR has a regional section, and the EMEA figures (8,245 incidents, 6,060 confirmed breaches) are sharper than the global ones:
- exploitation of vulnerabilities is the initial access vector in 47% of EMEA breaches, phishing in 28% and credential abuse in 6%;
- third parties are involved in 54% of EMEA breaches;
- the human element is present in 70%;
- System Intrusion accounts for 57% of breaches, up from 53%;
- state-affiliated actors appear in 23% of EMEA breaches against 14% overall, and espionage motives in 27% against 13%.
Verizon notes that its dataset has historically leaned towards North America and that it has worked to widen regional coverage. Even with that caveat, the message for a European IT manager is clear: the exposed perimeter matters even more here than in the global average.
Third parties: 48% and rising
Breaches involving a third party rose by 60% in a year, after doubling the year before. The report's survival analysis of third-party cloud exposures explains part of the problem: only 23% of third-party organisations fully remediated missing or poorly configured MFA on their cloud accounts, and for weak passwords and permission misconfigurations it took almost eight months to resolve half of the findings.
For NIS2 entities this maps directly onto Article 21(2)(d), supply chain security. The practical lesson is to ask critical suppliers for dated evidence rather than questionnaires. We looked at two 2026 cases in detail in breaches through suppliers: Inditex and Naturgy, and the regulatory side in NIS2 supply chain security.
Ransomware: more frequent, less paid, mostly small
- Ransomware was present in 48% of breaches, up from 44%.
- 69% of ransomware victims did not pay, and the median ransom paid fell to USD 139,875 from USD 150,000.
- Where the organisation's size was known, about 96% of ransomware victims were small and medium-sized businesses, which the DBIR defines as fewer than 1,000 employees. Verizon describes many of these attacks as opportunistic: victims were hit because they had compromised credentials (38%) or unpatched vulnerabilities in edge devices (29%).
- Of ransomware victims that had an earlier infostealer or credential leak, half had that event within the 95 days before the attack.
- In offers from initial access brokers, 44% of the access types sold were VPN access.
For the SMB segment as a whole (7,152 confirmed breaches), the initial access vectors were exploitation of vulnerabilities (26%), credential abuse (13%) and phishing (9%), and a third party was involved in 55% of breaches. The idea that smaller companies are not interesting to attackers does not survive these numbers.
What did not change: people
The human element, meaning an error, a misused privilege, stolen credentials or social engineering, is present in 62% of breaches. Social Engineering is the third most common pattern at 16%; phishing stays at 16% as an initial access vector and pretexting reaches 6%. Two shifts matter for training programmes:
- in phishing simulations, the median click rate through mobile channels such as voice and text messages is 40% higher than through email;
- 45% of employees are now regular users of AI tools on corporate devices, up from 15%, and 67% of those using AI services do so with non-corporate accounts. Shadow AI is now the third most common non-malicious insider action in Verizon's data loss prevention data.
On attackers' use of generative AI, the report is measured: the median threat actor researched or used AI assistance across 15 documented techniques, mostly to speed up well-known methods rather than invent new ones.
Reading the DBIR correctly
The DBIR is the most cited breach study in the industry, and also one of the most misquoted. A few points help to use it well:
- It is a sample, not a census. The data comes from Verizon's investigations and from dozens of contributing organisations. Percentages describe the breaches in that dataset, and the report itself notes its historical tilt towards North America.
- Percentages have different bases. The 31% for exploitation refers to breaches that were not errors or misuse and had a known entry point; the 48% for ransomware refers to all breaches. Quote the figure with its base.
- Definitions change between editions. This year's reclassification of pretexting is one example. Year-on-year comparisons are only meaningful where the report says the measure is comparable.
- It describes 2025, not today. Incidents ran to October 2025. Treat it as a direction of travel and measure your own organisation against it.
What it means for NIS2 entities
Each headline finding maps onto one of the measures in Article 21(2) of the NIS2 Directive, which is a useful way to turn the report into a work plan:
| DBIR 2026 finding | NIS2 Article 21(2) measure |
|---|---|
| Exploitation at 31% (47% in EMEA); KEV remediation at 26% | (e) vulnerability handling and disclosure |
| Third parties in 48% of breaches | (d) supply chain security |
| Credential abuse somewhere in 39% of breaches | (i) access control and (j) multi-factor authentication |
| Human element in 62%; mobile lures more effective than email | (g) basic cyber hygiene and training |
| Ransomware in 48% of breaches | (b) incident handling and (c) business continuity and backup |
Five actions for a European mid-sized company
- Know what you expose, continuously. You cannot prioritise exploited vulnerabilities on internet-facing systems if you do not know which systems face the internet. Inventory from the outside, not from the CMDB.
- Measure your own KEV time. Compare your median time to fix known-exploited vulnerabilities on exposed assets with the DBIR's 43 days, and set a target in days for those, not for everything.
- Ask suppliers for evidence. Start with the ones that hold your customers' data or can reach your network: MFA coverage, a dated external assessment, a named incident contact.
- Cover the gap with detection. Some exploitation happens before any patch exists. Someone has to be watching the VPN, the identity provider and the endpoints when it does.
- Protect credentials and train for the phone. Phishing-resistant MFA for administrators, monitoring for leaked credentials, and awareness training that includes voice and SMS pretexts.
Where Orizon fits
The DBIR's headline finding is a perimeter problem. RECON External maps every subdomain, port and service you expose, matches detected products to known CVEs weighed by CISA KEV, Exploit-DB, Metasploit and EPSS, and tests each finding from an isolated server, marking it Proven, To verify or Closed with the evidence. RECON finds it; Fireline proves it, running an autonomous penetration test on the external perimeter you authorise, to show what an attacker could actually do. For the gap that patching cannot close, Oversight, Orizon's SOC as a Service on the CYREBRO platform, reads the security tools you already have around the clock, with analysts on duty 24/7.
The 2026 DBIR does not ask for anything exotic. It asks companies to know their perimeter, fix what is exploited on it first, hold suppliers to evidence and keep watching. If you want to start from the first of those, run RECON External on your domains and see what is proven.