Small and mid-sized businesses have treated security as a big-company problem, something for banks and retailers with security teams and millions of customer records. For years that was a fair call. Breaking into software took skilled people and weeks of work, so attackers spent their time on the biggest targets, and serious protection cost more than a smaller company could justify.
Neither is true now. Attacks are automated, and they are aimed at everyone at once. In the Verizon 2025 Data Breach Investigations Report, ransomware was a component of 88% of breaches at businesses with fewer than 1,000 employees, and 39% at larger ones.
And it is speeding up. In November 2026, Google reported the first attacks it has seen in which AI turned a published fix into a break-in with no person involved. This article lays out what the evidence says about small and mid-sized businesses: who gets hit, how attackers get in, why the time to respond has shrunk to days, and what it costs.
Smaller companies aren’t overlooked
Large companies have security teams and budgets, so attackers go where those are missing. The Verizon 2025 Data Breach Investigations Report put it this way: “In larger organizations, Ransomware is a component of 39% of breaches, while SMBs experienced Ransomware-related breaches to the tune of 88% overall.” Verizon’s SMBs are businesses with fewer than 1,000 employees. Ransomware locks a business out of its own systems and data until someone pays.
The victims are mostly smaller companies, and there are more of them every year. Ransomware groups publish the names of the companies they hit, and Black Kite counts them. In the year to March 2026 it counted 7,551, up from 6,046 the year before, and 49.3% were in the United States. Among the victims whose revenue is known, companies above $100 million made up 9.5%. The largest group, at 37.4%, had revenue between $10 million and $50 million.
Ransomware lands on smaller companies
Two measures
- Small and mid-sized businesses, fewer than 1,000 employees: ransomware was a component of 88% of breaches
- Larger organizations, 1,000 employees or more: ransomware was a component of 39% of breaches
- Named ransomware victims with revenue under $10 million: 23.9%
- Named ransomware victims with revenue $10 million to $50 million: 37.4%
- Named ransomware victims with revenue $50 million to $100 million: 29.3%
- Named ransomware victims with revenue over $100 million: 9.5%
- 7,551 companies named as ransomware victims in the year to March 2026, up from 6,046; 49.3% in the United States
The money lost has grown with it. In 2019, people and businesses in the United States reported $3.5 billion in losses from internet crime to the FBI. In 2025, they reported $20.877 billion.
Losses from internet crime reported to the FBI
United States, 2019 to 2025
- 2019: $3.5 billion reported lost
- 2020: $4.2 billion reported lost
- 2021: $6.9 billion reported lost
- 2022: $10.3 billion reported lost
- 2023: $12.5 billion reported lost
- 2024: $16.6 billion reported lost
- 2025: $20.877 billion reported lost
Smaller companies are a large part of that. Microsoft surveyed 2,000 businesses with 25 to 299 employees in the United States and the United Kingdom, and one in three had been through a cyberattack in the past year. Sophos investigates break-ins for its customers. In 2023, more than three in four of those cases were at small businesses.
One case shows how it happens. In September 2025, Microsoft tracked a single ransomware group as it hit more than 50 organizations in four weeks. Most were small and medium-sized businesses, and the attacks were tied to one flaw in a VPN product, published the year before.
They get in through software nobody updated
Verizon studies thousands of real breaches every year and records how each one started. For years, the most common way in was a stolen password. That has changed. In the Verizon 2026 Data Breach Investigations Report, the most common way in is a known flaw in software. In the 2024 report, 14% of breaches began that way. In 2025 it was 20%. In 2026 it is 31%, and stolen passwords are down to 13%.
The same is true at smaller companies. At businesses with fewer than 1,000 employees, 26% of breaches began with a software flaw and 13% began with a stolen password. More and more, the thing that gets breached is a web application. In the 2026 report, 54% of breaches involved one. A year earlier it was 43%.
How breaches start
2024 to 2026 reports
- 2024 report: 14% of breaches began with an exploited vulnerability
- 2025 report: 20% of breaches began with an exploited vulnerability
- 2026 report: 31% of breaches began with an exploited vulnerability
- Small and mid-sized businesses, 2026 report: Exploitation of vulnerabilities, 26%
- Small and mid-sized businesses, 2026 report: Credential abuse, 13%
- Small and mid-sized businesses, 2026 report: Phishing, 9%
- 2025 report: a web application was affected in 43% of breaches
- 2026 report: a web application was affected in 54% of breaches
“Being a small organization, you may mistakenly think that your threat profile and who would be interested in compromising you are significantly different from everyone else. That is, until you find yourself on the wrong side of a ransomware attack.”
Verizon isn’t alone in finding this. Other firms that investigate break-ins publish how each one began, and they count their own cases. Google’s Mandiant team has ranked exploited software first for six years running, at 32% of the intrusions it investigated in 2025. IBM’s responders put it at 40% of their incidents, ahead of attackers logging in with valid credentials at 32%. In Microsoft’s incident response work it rose from 15% of cases to 24% in a year.
Nobody has to choose your company for this to happen. It’s natural to assume a smaller business isn’t worth an attacker’s time, and that would be true if a person were picking the targets. Mostly, software does it. Automated tools try every login page, form, and API they can reach, whoever owns it.
Thales, a security company that filters this traffic for the websites it protects, measures it every year. In 2015, bad bots made up 15% of all internet traffic. In 2025, they made up 40%, two in every five visits. Akamai runs one of the largest networks that deliver and protect websites, so it sees the attacks arrive. It counted an average of 258 API attacks a day against each of its enterprise customers in 2025, up from 121 the year before.
Automated attacks
2015 to 2025
- 2015: bad bots were 15% of internet traffic
- 2016: bad bots were 20% of internet traffic
- 2017: bad bots were 22% of internet traffic
- 2018: bad bots were 20% of internet traffic
- 2019: bad bots were 24% of internet traffic
- 2020: bad bots were 26% of internet traffic
- 2021: bad bots were 28% of internet traffic
- 2022: bad bots were 30% of internet traffic
- 2023: bad bots were 33% of internet traffic
- 2024: bad bots were 37% of internet traffic
- 2025: bad bots were 40% of internet traffic
- 2024: 121 API attacks per company per day
- 2025: 258 API attacks per company per day
The Verizon 2026 Data Breach Investigations Report describes ransomware groups as “casting out wide nets to compromise as many organizations as possible,” and says it’s “not so much about industry or the revenue of the victims.” What gets an app hit is being reachable and unpatched when a sweep comes through. The sweeps keep coming.
A published fix starts a clock
Very little of this is about the code written for your app. Every app is built on other people’s software: a programming language, a framework, a database, a server, and open-source packages by the hundreds. Black Duck audits commercial codebases for companies that are buying or investing in them. In its 2026 report, the average app contained 1,180 open-source components, up from 911 a year earlier.
When someone finds a security flaw in any of that software, the people who maintain it fix it, and the flaw is published in a public catalog along with the fix. This is the system working as it should. People all over the world do this work responsibly, so that everyone who uses the software knows to update.
The same announcement tells attackers where to look. They compare the old version with the fixed one, and the difference shows them where the hole was. Then they go looking for every app that hasn’t installed the fix.
How a published fix becomes an attack
Four steps
Working out an attack from a fix used to take skilled people weeks, and that gave everyone else time. Google’s Mandiant team measures how long it takes, on average, for a newly public flaw to be used in an attack, across every kind of software. In 2018 and 2019 it was 63 days. By 2023 it was 5.
Days from a flaw going public to its first attack
Average, by period
- 2018 to 2019: 63 days
- 2020 to early 2021: 44 days
- 2021 to 2022: 32 days
- 2023: 5 days
- Rapid7, third quarter 2026: 3 days. The average from a serious flaw becoming public to its exploitation.
A business that updated its software once a month used to be ahead of most attacks. Now the attacks arrive first. Rapid7 tracks the gap too, and in the third quarter of 2026 its average for serious flaws was three days from going public to being exploited. Nine in ten of the attacks it tracked came within 30 days. Google recorded 141 flaws that were disclosed and exploited in the first eight months of 2026, more than in all of 2025.
Now AI does it on its own
What changed is AI. Comparing two versions of a program to find what changed is something it does well. In 2026, Anthropic gave its most capable model 18 recent security fixes for the Firefox browser. It wrote its first working attack in under an hour, and eight within about 12 hours. The researchers’ conclusion: “A lone operator can now turn a month’s worth of patches into working exploits in a single afternoon.”
Attackers are already past needing the operator. On November 16, 2026, Google’s Threat Intelligence Group reported that, for the first time, it had found attackers using AI to find and exploit flaws with no person in the loop. The tool watched popular web application software, and the server it ran on held data taken from its victims.
This is the part to take seriously. Until now, each of these attacks needed a skilled person, and there are only so many of those. Software has no such limit. It can work through every new fix, against every site it can reach, around the clock. Google’s conclusion is that attackers who chain these steps together “can dramatically reduce the window between patch disclosure and weaponized exploitation.” The US government’s cyber agency now tells federal agencies that attackers’ use of AI “may further narrow the time defenders have to react between patch release and possible exploitation.” We expect attacks like this to multiply fast.
There are also far more flaws to keep up with. In 2016, the National Vulnerability Database recorded 6,449 new flaws. In 2026, it passed 100,000 in November. That is more than fifteen times as many, and more than 2024 and 2025 put together. AI is part of that too. Anthropic’s Project Glasswing found thousands of unknown flaws in open-source code with it.
Why it keeps getting worse
Many of those flaws are in code your app already contains. Black Duck’s report covers 947 codebases. It found open-source code in 98% of them, a known vulnerability in 87%, and components more than four years out of date in 92%.
Almost none of that needs a new invention to fix. Sonatype tracks downloads of open-source packages. When a vulnerable package was downloaded, a fixed version already existed 95% of the time.
Old code inside apps
947 codebases audited
- Commercial codebases that contain open-source code: 98%
- Commercial codebases that use components more than four years out of date: 92%
- Commercial codebases that contain a known vulnerability: 87%
- Vulnerable package downloads where a fixed version already existed: 95%
So the risk in most apps isn’t a missing fix. It’s that updating is nobody’s job. The team that built the app moved on to the next project, the updates pile up, and each one gets harder to install the longer it waits.
No language or framework is exempt. Datadog measures the services its customers run in production. Of those services, 40% had an exploitable vulnerability, from 59% on Java to about 12% on Ruby, and one in ten ran on a language version past its end of support. Whatever your app is built on, it needs looking after.
Every stack has the problem
Production services, by language
- Java: 59% of services had an exploitable vulnerability
- .NET: 47% of services had an exploitable vulnerability
- Rust: 40% of services had an exploitable vulnerability
- All languages: 40% of services had an exploitable vulnerability
- JavaScript: about 37% of services had an exploitable vulnerability
- Go: about 35% of services had an exploitable vulnerability
- Python: about 27% of services had an exploitable vulnerability
- PHP: about 23% of services had an exploitable vulnerability
- Ruby: about 12% of services had an exploitable vulnerability
- Go: 23% of services ran a version past its end of support
- PHP: 13% of services ran a version past its end of support
- Java: 10% of services ran a version past its end of support
- All languages: 10% of services ran a version past its end of support
- Node.js: about 9% of services ran a version past its end of support
- Python: about 8% of services ran a version past its end of support
- Ruby: about 5% of services ran a version past its end of support
Fewer companies have anyone whose job this is. The clearest count comes from Canada’s national statistics agency, which surveys about 12,000 businesses and asks how many employees work on security as a main part of their job. In 2019, 57.7% of businesses with 10 to 49 employees had at least one. In 2025, 44.5% did. Among those with 50 to 249, it fell from 67.3% to 51.1%. In the United States, Guardz surveyed 800 small and mid-sized businesses in 2025. In a third of them, the owner personally handles security alerts.
Businesses with anyone working mainly on security
By number of employees
- Large businesses, 250 or more employees: 2019, 85%; 2021, 86.1%; 2023, 82.8%; 2025, 73.6%
- Medium businesses, 50 to 249 employees: 2019, 67.3%; 2021, 66.6%; 2023, 60.1%; 2025, 51.1%
- Small businesses, 10 to 49 employees: 2019, 57.7%; 2021, 58.8%; 2023, 47.6%; 2025, 44.5%
The work is growing, and fewer companies have anyone assigned to it.
Why regular maintenance doesn’t cover this
Plenty of businesses already pay someone to maintain their app, and it’s fair to ask why that doesn’t cover security. Most maintenance agreements buy time for changes and a round of updates on a schedule, often once a month. That was a sound arrangement when attacks took two months to arrive.
What it rarely includes is anyone watching. An app depends on a server, an operating system, a database, a language, a framework, and hundreds of packages, and new flaws are published in them every day. Under most agreements, nobody is responsible for checking each one against your app the day it appears. Nobody is on a deadline to fix the ones that matter.
So a flaw published in the first week of the month waits for the next round of updates. The app sits open for three weeks, and the attacks arrive in the first.
How long an app stays open after a flaw is published
Days
- First attacks arrive 3 to 5 days after a flaw is published, on average
- Web apps and APIs, Average time to fix a serious flaw: 54.8 days
- The typical business, To fix a flaw already under attack: 43 days
- A monthly round of updates, Up to a month, for a flaw published just after one: 30 days
- Essential, Critical security fixes: 7 days
- Proactive, Critical security fixes: 3 days
- Critical care, Critical security fixes: 24 hours
Most teams are slower than monthly. Edgescan tests web apps and APIs for a living, and it finds that serious flaws stay open for 54.8 days on average. The Verizon 2026 Data Breach Investigations Report looked only at flaws already under attack, and the median time to fully fix them was still 43 days. Veracode finds that half of all application flaws are still open after 243 days, and half of the flaws in open-source libraries after 358.
The fix itself is usually small. Installing an update can take an engineer less than an hour. The hard part is knowing, the day a flaw is published, that it affects your app, and having someone whose job is to install the fix that week. That is what most businesses are missing, and it’s what to look for in any arrangement. Three questions will tell you whether you have it.
When a serious flaw is published in something we use, who finds out, and how fast?
“We’d hear about it” is a hope. A good answer names a person or a system that watches every part of the app 24/7 and is alerted the moment a new flaw is published. Checking once a day already gives attackers a day’s head start.
How long did our last critical security fix take, from published to installed?
If nobody can say, nobody is measuring. A good answer is a number of days, with a record to back it up.
Who is responsible for installing fixes, and by what deadline?
“The developers get to it when they have time” means security competes with every feature request. Look for a named owner and a deadline in writing.
What it costs
Coalition insures businesses against cyber incidents, and it publishes what its claims cost. Across every claim made in 2025, the average loss was $116,000. For companies with under $25 million in revenue it was $77,000. Between $25 million and $100 million it was $170,000, and above $100 million it was $268,000.
The average cyber insurance claim
By company revenue
- All claims: $116,000 average claim
- Under $25M in revenue: $77,000 average claim
- $25M to $100M in revenue: $170,000 average claim
- Over $100M in revenue: $268,000 average claim
Coalition’s own note on the smallest companies: “a $77,000 loss can be far more disruptive to a micro SMB balance sheet than a six-figure loss is to a large enterprise.” And a claim only counts what insurance paid. It leaves out the downtime, the customers who left, and the weeks a team spent on nothing else. When Microsoft asked smaller businesses to add all of that up, the average total cost of an attack came to $254,445.
What a plan does about it
The Black Airplane Protection Plan is our answer to those three questions. If no one at your company works on security, our team does that work for you. Our senior engineers watch every part of your app around the clock, install critical security fixes by a written deadline, and update its packages every month. We also have plans that move the language, framework, and server to supported versions before the old ones lose support. You get a plain report of everything we did.
There are three plans, chosen by what’s at stake if the app were attacked. Most apps need Proactive. It is for any app that holds customer or business data, earns revenue, or ships to the app stores, and it installs critical fixes within three days. Critical care is for apps where a breach would do serious harm, like those holding health, financial, or payment data. Essential is only for sites and tools with nothing sensitive in them. A plan can cover an app we built or one another team built. For scale, the average claim above costs almost five years of our Proactive plan.
When you don’t need one
A plan like this isn’t for everyone. If you have in-house engineers whose job includes patching, with a deadline they’re held to, you already have what a plan provides. If your site runs entirely on a hosted platform like Shopify or Squarespace, with no custom code, the platform does this work for you. And if an app is being retired in the next few months, the money is better spent on its replacement.
Everyone else is in the position the numbers describe: an app that matters to the business, built on software that changes every week, with nobody assigned to keep up.
Frequently asked questions
Why are so many more software vulnerabilities being published?
More software exists, more people are looking for flaws in it, and AI now helps them look. The National Vulnerability Database recorded 6,449 new flaws in 2016 and more than 100,000 in 2026, more than fifteen times as many. For a business, the problem is volume. Someone has to check each new flaw against the app and install the updates that apply.
Do attackers really go after small and mid-sized businesses?
Yes, because most attacks don’t choose a target at all. Automated tools test every app they can reach, and bad bots now make up 40% of internet traffic, according to Thales. In the Verizon 2025 Data Breach Investigations Report, ransomware was a component of 88% of breaches at small and mid-sized businesses and 39% at larger organizations. Being small doesn’t hide an app from a scan.
We already pay for a maintenance plan. Doesn’t that cover security?
Usually not. Most maintenance agreements buy hours for changes and a round of updates on a schedule, often monthly. Few make anyone responsible for watching every part of the app 24/7 and fixing a serious flaw within days. A flaw published just after a monthly round can stay open for weeks. Ask who finds out when a flaw is published, and what deadline they work to.
How fast does a critical security fix need to be installed?
Within days. Google’s Mandiant team found that attacks on a newly public flaw began 5 days after it was published, on average, in 2023. Most teams take far longer. Serious flaws in web apps and APIs stay open for 54.8 days on average, according to Edgescan. Our Protection Plan installs critical security fixes within 3 days on Proactive, the plan most apps need, and within 24 hours on Critical care.
Doesn’t our hosting company handle security updates?
Your host patches its own platform. Your app, its packages, and often its server are yours to keep current. Even the best hosting agreements exclude anything your app causes. If you aren’t sure where the line sits, ask your host which updates it installs and which it leaves to you. Then ask who on your side is installing the rest.
Can a protection plan guarantee we won’t be hacked?
Nobody honestly can, and it’s worth being wary of anyone who says otherwise. What a plan changes is how long your app stays exposed. We watch around the clock, fix by the deadlines in your plan, keep your app current, and report every outage and its cause. Security stops being something nobody owns and becomes work that gets done on a schedule.
Put a deadline on it
The evidence points one way. Smaller companies are well within reach of these attacks, and most of them arrive through software nobody is keeping current. Attackers now turn a published fix into an attack in days. Most apps get that fix weeks later. The businesses best placed are the ones where someone watches for new flaws every day and installs the fixes by a date.
If that isn’t true of your app yet, the Black Airplane Protection Plan is how we make it true. Every plan starts with an assessment of where your app stands today.



