Written by: Mio Kanzaki(Security Analyst (AI))Reviewed by: 千葉圭介Updated
Data leak prevention means making sure that personal data, documents entrusted to you by partners and internal secrets do not reach people they were never meant for. It concerns operators of sites that hold inquiry or member data, companies with shops or booking systems, and every company that sends personal data by email or on paper. Three jobs come first: take stock of the personal data you hold, strengthen admin screens and logins, and set a rule to check the recipient before anything is sent.
This page is a hub for executives and site owners. It shows what leaks and from where, what to do first, and when a report or a notice is needed. If you already suspect a leak, read the emergency guide first.
What counts as an information leak, and where does data leave?
Many people picture an outside attacker carrying off a customer list. The public figures below point elsewhere: in the reports that Japan's Personal Information Protection Commission received directly in fiscal 2025, wrongly delivered documents and misdirected sending made up most of the count.
Data leaves in two ways. One is malicious: unauthorized access from outside, or misuse by a current or former employee. The other is accidental: documents handed to the wrong person (wrong delivery), an email or letter sent to the wrong address (misdirected sending), a lost item, or a disposal mistake. Accidents call for procedures and tooling. Malicious acts call for closed entrances and a way to notice data being taken.
What causes leaks, and where does data get out?
Knowing the route helps you decide what to fix first. The table sorts the routes a site owner should keep in mind by type.
| Route | What happens | Typical entrance | Read more |
|---|---|---|---|
| Wrong delivery and misdirected sending | A document or email reaches the wrong person | Post, bulk email, handing over at a counter | Item 10 of the "10 things to do today" below |
| Flaws in how a web application is built | A database is read, or one customer sees another's data | Contact forms, search boxes, member pages | SQL injection, broken access control |
| Public exposure through a setting | Private files can be seen by anyone | Server settings, cloud storage, backup locations | Security misconfiguration, cloud storage exposed by mistake |
| Leaked credentials and keys | Someone enters as a legitimate user and reads data | Admin screens, cloud keys, secrets in source code | Access key leak |
| Card data skimmed from an online shop | Card numbers go elsewhere as a buyer types them | Altered checkout pages, e-commerce parts not updated | Card skimming (web skimming) |
| Leaks at partners and contractors | Data you entrusted leaves through a contractor's incident | A contractor's servers, maintenance permissions | Preventing supply chain attacks |
| Input into generative AI or unapproved services | Material pasted by staff stays in an outside service | Cloud and AI use nobody has mapped | LLM data leak, shadow IT |
From the second row down, the entrance is a hole someone found, a setting that went wrong or a permission that was handed over. We do not publish the steps or strings that attackers actually use.
What do public statistics show?
Figures give you a yardstick for your own situation. We opened the original publications to check each number below.
Personal Information Protection Commission: 17,139 reports in a year, and two-thirds of the direct ones are wrong delivery or misdirected sending
In its annual report for fiscal 2025 (April 2025 to March 2026), the Commission handled 17,139 reports of leaks and similar incidents at private businesses, against 19,056 the year before. It received 13,345 of them directly. The other 3,794 came through ministries with delegated authority.
The breakdown by cause covers the 13,345 direct reports. Where the reporting business itself was the source, wrong delivery accounted for 6,072 cases (45.5%) and misdirected sending for 2,889 (21.6%), about two-thirds of the 13,345 together. Lost items came to 654 (4.9%), unauthorized access to 533 (4.0%) and insider misconduct to 75 (0.5%).
Unauthorized access where a contractor was the source added another 328 cases (2.5%). An incident at a service provider can expose what you entrusted, even when your own defenses held.
Sorted by the situation that triggers a reporting duty, the 17,139 reports look like this: special care-required personal information in 10,603 (61.9%), a suspected unlawful purpose in 3,822 (22.3%), risk of financial harm in 3,525 (20.6%) and more than 1,000 people concerned in 883 (5.2%). Some cases meet more than one test, so the total exceeds 100%.
For unauthorized access, the Commission names three causes: a published vulnerability in a VPN device or in the application behind an online shop that the business left unfixed although a remedy existed, easily guessed IDs and passwords, and a setting mistake that left database access control wrong. Checking these entrances is part of the defense, next to reviewing human error in daily work.
IPA: attacks on confidential information and insider misconduct are in their 11th year on the list
In the Top 10 Information Security Threats 2026, released by IPA on 29 January 2026, "targeted attacks aimed at confidential information" is fifth among threats to organizations, and "information leaks through insider misconduct" is seventh. Both have been selected for 11 consecutive years. Second is "attacks aimed at the supply chain or contractors", selected for eight consecutive years, and leaks through a contractor fall under it.
Third place, "cyber risk from the use of AI", is new this year. Staff pasting business material into a generative AI service is one more route to map.
Cases around online shops and payment
Card records leaked in about 140,000 cases from EC-CUBE sites, according to a 2019 report by the Ministry of Economy, Trade and Industry (IPA alert of 25 December 2019). JVN published a vulnerability in the official Amazon Pay plugin on 14 October 2026 (CVE-2026-103671), and the developer has confirmed that it was exploited. Do not postpone updates to payment plugins.
Once you have seen the figures, you may want to know where your own site stands. The free URL check reads only what anyone can see from outside, and it costs JPY 0.
When must you report a leak, and when must you tell the people concerned?
A company may have to report a leak and notify the people affected. Below are the main points of the Commission's guidance under the Act on the Protection of Personal Information. In borderline cases, read that guidance or ask a specialist.
Four situations that require a report
The guidance covers leaks, loss or damage of personal data, and the risk of them. A report is required in these four kinds of situation:
- The data includes special care-required personal information, such as medical history.
- There is a risk of financial harm, for example card numbers, or the ID and password of a service that can transfer money.
- The act is suspected to have an unlawful purpose, such as unauthorized access, theft, removal by an employee or website tampering.
- The number of people concerned exceeds 1,000.
A report has two stages. The guidance asks for a prompt preliminary report once you find out, with 3 to 5 days from discovery as a guide. The final report follows within 30 days of discovery, or 60 days if an unlawful purpose is suspected. If the incident happened at a contractor, the business that entrusted the data may still be the one that must report.
How do you think about notifying the people concerned?
Separately from the report, the people affected must be told. The Commission's guidelines say to notify them promptly once you learn of a covered situation. The notice gives an outline, the items that leaked, the cause, whether secondary harm is possible and any helpful information. You need not wait until everything is known. If you cannot reach people, a substitute is accepted: a public announcement, or a help desk where people can ask whether they are affected. Keep records as you go.
What should you do first, by company size?
Pick the steps that match your company.
A company of a few people up to about 20, with site updates outsourced
Staff errors and a site nobody updates can both be the starting point. Work in this order: list the personal data you hold, set a rule for checking email addresses, tidy administrator accounts and turn on two-factor authentication.
If you outsource updates, ask in writing where the personal data is held. A vague answer means there are parts you cannot see yet, so keep asking.
A company of a few dozen with several sites or systems
Check cloud storage and backup locations for files that should not be public. Permissions come first: list who can see which data, remove access for people who left or changed roles, and add contractors' connection routes to the same list. Check what providers must tell you if a leak occurs.
A company with an online shop, member site or booking system
You handle card and member data directly, so a leak may meet the reporting tests, such as risk of financial harm. Entrances to watch are altered checkout pages, e-commerce parts not updated and weak authorization on screens behind login. Build three things into operations: an assessment before launch and after each large change, detection of changes to checkout pages, and an emergency contact with the person who decides on reports.
10 things to do today
These 10 steps reduce the entrances for a leak and fix in advance what to do when one happens. They combine work from our fix menu with operating rules. Some items assume WordPress, but the idea is the same on other CMSs.
- List the personal data you hold. Write down what you hold, where, and how many records, including saved form entries, copies in email, backups and paper. Data whose location you do not know cannot be brought into your protection.
- Apply updates and remove unused parts. Bring the core, themes and plugins up to date, and delete anything merely deactivated.
- Tidy administrator accounts and passwords. Delete administrators nobody needs, and end weak and reused passwords. Check for accounts of people who left and of earlier agencies.
- Turn on two-factor authentication. Even if a password leaks, nobody gets in without the phone confirmation (two-factor authentication).
- Restrict where admin screens can be opened from. Allow only fixed places. Ask us if you are unsure about server settings.
- Put sensitive files out of public reach and review permissions. Put configuration and backup files where outsiders cannot read them, and check that no backup or customer list sits in the public folder of the site.
- Stop entrances you do not use, and stop usernames from being visible. Close unused channels and make login names harder to guess.
- Detect file tampering. Set up something that notices when code to skim data is added to checkout pages or input forms.
- Encrypt traffic (SSL) and set security headers. Encryption keeps what visitors type in forms from being read on the way, and the headers switch on browser-side protection.
- Set a procedure against misdirected sending, and decide whom to contact when you notice a leak. Examples are BCC for bulk email and a second person checking the address before sending. Then write on one sheet who receives the report and which executive decides whether to notify the authorities.
Begin with items 1 to 4. Take a backup before items 5 to 10, because a wrong setting can lock you out of the site. Finishing all 10 does not guarantee that no leak occurs.
Short of time to do it yourself? Tell us your situation and we will go through which of the 10 are still open.
What should you hand to a specialist?
Ask two questions before deciding who does a job: can a mistake be undone, and does the job need skill to examine the inside of the site?
We recommend a specialist for these:
- Vulnerability assessment. Looking from outside cannot show whether your site has a hole that exposes data. A WordPress/CMS site assessment is from JPY 49,800 (excl. tax, up to 30 pages). One re-assessment after your fixes is free. Screens behind login and paid plugins both change the price.
- Member, payment and API functions, where a leak would be costly. A web application assessment Lite is from JPY 98,000 (excl. tax, up to 10 screens). The Standard tier is from JPY 198,000 (excl. tax). It covers up to 20 screens and manual checks that include screens behind login. An API assessment is from JPY 98,000 (excl. tax, up to 10 endpoints).
- Cloud permissions and public settings. A cloud settings assessment for AWS, Google Cloud or Azure is from JPY 98,000 (excl. tax, per account).
- Investigation and recovery when a leak is suspected. If you touch files first, you can wipe out the evidence, so follow an order of steps. Recovery from tampering and malware damage is from JPY 29,800 (excl. tax). A same-day start adds JPY 20,000 (excl. tax).
- Continuous watch and answers to partners. Maintenance Lite is JPY 9,800 per month (excl. tax), and Guard Lite, our AI monitoring for WordPress, is JPY 5,000 per month (excl. tax). The security advisory plan is JPY 29,800 per month (excl. tax) and helps with check sheets.
We fix holes found in an assessment, investigate and recover after damage, help prepare reports and keep watch afterwards. Reporting to the Commission and notifying the people concerned are procedures the business carries out, and we do not file them for you. We help sort out what happened and prepare material for the report.
Before delivery you receive a PDF of every change and check the real screens yourself. For the first month after delivery, support costs nothing. We cannot promise that a leak will not happen. An assessment confirms that the holes known now are absent, and says nothing certain about the future.
Which pages cover the individual holes?
Every page below covers the mechanism of a hole linked to data leaks, how to check it and the fix.
Web applications and APIs
- SQL injection
- Broken access control
- Path traversal
- Server-side request forgery (SSRF)
- XML external entity (XXE)
- Security misconfiguration
- Cryptographic failures
- CORS misconfiguration
- Directory listing
- Information disclosure
- Broken object level authorization (BOLA/IDOR)
- Excessive data exposure in APIs
Smartphone apps
- Insecure data storage in smartphone apps
- Insecure communication
- Hardcoded secrets
- Reverse engineering
Cloud, e-commerce and CMS
- Cloud storage exposed by mistake
- Excessive IAM permissions
- Access key leak
- EC-CUBE vulnerabilities
- Shopify app and theme vulnerabilities
- Card skimming (web skimming)
- WooCommerce vulnerabilities
Generative AI, people and operations
- Prompt injection
- LLM data leak (a hole through which what staff or customers type in can come out in the AI's answers)
- Insider threat
- Shadow IT
For neighbouring topics, see hacking prevention, unauthorized access, website tampering and ransomware.
How can the free URL check help you decide?
Know your starting point before you choose what to fix. Enter your site's address in the free URL check to see where you stand. It reads only what anyone can see from outside: the page you submit, public DNS and the encrypted connection. It looks at the certificate expiry, security headers, software versions that appear on the page, and SPF and DMARC settings, and it compares the JavaScript libraries the page loads against known vulnerabilities. It sends no attacks and cannot see screens behind login or the inside of your server. It costs JPY 0.
Where the result raises a concern, we can sign an NDA, take your assessment consent form and run the free assessment, a quick check by tools plus a person. You see the issue report before deciding on fixes.
Frequently asked questions
Is hacking the most common cause of data leaks?
Not by count. In the Personal Information Protection Commission's figures for fiscal 2025, the Commission received 13,345 reports directly. Among those where the reporting business itself was the source, wrongly delivered documents were 45.5%, misdirected sending 21.6% and unauthorized access 4.0%. Work on both staff procedures and system defenses.
Do we always have to report a leak to the Personal Information Protection Commission?
Not always. A report is required when the leak involves special care-required personal information, may cause financial harm, was done with an unlawful purpose, or concerns more than 1,000 people. Whether you are covered depends on the case, so read the Commission's guidance or consult a specialist.
Can personal data leak from a small company's website?
It can. A small site stores personal data too, such as saved contact-form entries, member lists and booking records. The Commission names three causes of unauthorized-access leaks: published vulnerabilities left unfixed, easily guessed IDs and passwords, and database access control left wrong by a setting mistake.
How much does data leak prevention cost?
The free URL check, which looks only at what is visible from outside, costs JPY 0. A WordPress/CMS site assessment is from JPY 49,800 (excl. tax, up to 30 pages). A web application assessment Lite is from JPY 98,000 (excl. tax, up to 10 screens). The price depends on pages, screens behind login and paid plugins. A formal quotation follows the free assessment.
Next step
Do two things today: work through items 1 to 4, and run your site through the free URL check.
To talk through priorities and what to outsource, contact us. If you already suspect a leak, begin with the emergency guide.
Written by Mio Kanzaki (AI security analyst). Supervised by 千葉圭介. Updated 19 November 2026.
Frequently asked questions
Is hacking the most common cause of data leaks?
Not by count. In the Personal Information Protection Commission's figures for fiscal 2025, the Commission received 13,345 reports directly. Among those where the reporting business itself was the source, wrongly delivered documents were 45.5%, misdirected sending 21.6% and unauthorized access 4.0%. Work on both staff procedures and system defenses.
Do we always have to report a leak to the Personal Information Protection Commission?
Not always. A report is required when the leak involves special care-required personal information, may cause financial harm, was done with an unlawful purpose, or concerns more than 1,000 people. Whether you are covered depends on the case, so read the Commission's guidance or consult a specialist.
Can personal data leak from a small company's website?
It can. A small site stores personal data too, such as saved contact-form entries, member lists and booking records. The Commission names three causes of unauthorized-access leaks: published vulnerabilities left unfixed, easily guessed IDs and passwords, and database access control left wrong by a setting mistake.
How much does data leak prevention cost?
The free URL check, which looks only at what is visible from outside, costs JPY 0. A WordPress/CMS site assessment is from JPY 49,800 (excl. tax, up to 30 pages). A web application assessment Lite is from JPY 98,000 (excl. tax, up to 10 screens). The price depends on pages, screens behind login and paid plugins. A formal quotation follows the free assessment.