The New Year’s-card program Fudegurume (筆ぐるめ), or the Pocketalk translator, can sit unused for years. An email address and one easy password stay on the membership page. The same password still opens the mailbox, a shop, or a work account. After the 15 October notice, the usual ending is two sentences: the password was stored encrypted, and this site’s old passwords are already dead, so the other sites can wait. Those two sentences each drop half of what was published. The notice itself never uses the word encrypted. The sentence that the passwords left already decrypted is the 19 October interview. The sentences that kill this site’s old password sit on the FAQ and on that interview. None of them edits the login box on a shop.
This page answers one question: after a software site says the decryption key never left, but the passwords left already decrypted, and this login has been voided, where do the email, the phone number, the company title, and the same password on other sites still sit. The facts follow SOURCENEXT’s 15 October 2026 notice, its FAQ on the same incident, Nikkei’s 15 October story (updated at 00:19 on 16 October), NHK’s 16 October story, and ITmedia’s 19 October interview by 岡田有花. This page does not explain how to call the member-information API. MyPassGen’s generator and strength check open without an account. A password you type stays in the browser.
Do the verb the notice already uses
On 15 October they asked anyone who still uses the same email address and password on another service to change the password there, because a third party may try that combination. They also asked you to leave links and attachments in suspicious mail, SMS, calls, and postal items unopened, and to keep personal information and passwords off a page you do not trust. The same notice says that, at login, you are asked to reset this account and to follow the on-screen instructions promptly. The FAQ later says a sudden failure to sign in is their invalidation, and that invalidating the passwords included in the leaked set is complete. ITmedia on 19 October says they force-reset every customer’s password. Those sentences are about this site’s login box. A copy that already left in decrypted form stays outside.
Voiding this login leaves the old password on other sites
Three published sentences describe what happened to the SOURCENEXT password, and they are not the same sentence. The 15 October notice says a measure is in place that asks you to reset the password when you log in, and asks you to follow the screen promptly. The FAQ answers “I suddenly cannot log in” with yes: to protect the account they are invalidating passwords in turn, so please set a new one. A later answer on that same FAQ says that, right after they found the intrusion, they fixed the weakness and completed the measure of invalidating the passwords included in the leaked set. ITmedia’s 19 October interview says the company force-reset every customer’s password and asked customers to set a new one. Read together, the old password at SOURCENEXT is the one they are killing. A shop that still accepts the same string does not read the notice, the FAQ, or the reset switch.
The FAQ holds both “in turn” and “the measure is complete.” This page leaves both sentences on the page that wrote them. The practical reading is narrow. Count on the old password to fail at SOURCENEXT, and if it still signs you in, the reset the notice asks for is still in front of you. A new password you manage to save on their reset page retires the next login there. It leaves the shop, the mailbox, and the work account on the old string until you change those yourself.
The reset page was hard to open while people rushed it. A 16 October report quotes a SOURCENEXT message: the password-reset site was under concentrated access and hard to reach, with an apology for the inconvenience. The same report says the reporter’s own attempt returned the error “password reminder processing failed” (パスワードリマインダ受付処理に失敗しました). That congestion message is not a paragraph of the 15 October notice. A crowded reset page means you may still be waiting to write a new password on this one site. The old password on a shop keeps working through the wait. Change the services you can change now. When the official reset page opens, set a password you have not used anywhere else. Those are two clocks.
The 19 October interview makes the new password more specific than “please reset.” Asked how passwords are managed after the reset, and whether they had moved to a hash, the spokesperson declined. The words ITmedia records are that the concrete management method and the detail of what they will do next are withheld to protect the security of the system. The new password you type on the reset page belongs on this site alone. Copying it back to the mailbox or the shop places a password, whose storage method the company refuses to describe, into another login box.
The window in the notice runs from about 10:00 on 9 October, when unauthorized use of the member-information retrieval API began, to 13:36 on 13 October, when containment finished. The notice does not list each pull by the hour. If you changed this site’s password during those four days, the published sentences do not prove the new password was taken, and they do not prove it was spared. Change the password on other sites that still matches the one in use before 9 October, which is the request they already made. After containment they wrote that further unauthorized access had not been confirmed. That is a monitoring conclusion at the time of the notice. A plaintext copy already outside does not expire because the monitoring line is quiet.
If, before 9 October, each site already used a different password, those login boxes do not hold this SOURCENEXT string. A decrypted copy of the old password can still sit outside. It simply fails on the sites you had already split. Contact details, the company name, and the job title do not depend on reuse. They stay on the field list either way.
Which fields the 15 October notice actually lists
SOURCENEXT (ソースネクスト) is a Japanese software company. Nikkei names the PC software, including the New Year’s-card program Fudegurume, and the Pocketalk translation device. Products are sold to individuals and to companies. Nikkei writes that cumulative membership exceeds 20 million. That figure is the size of the membership. The notice’s 22,519,960 is a count of records that may have left. The notice does not call it 22.52 million customers. Nikkei and later headlines round the count to about 22.52 million records (2252万件). Keep the record count and the membership figure in two sentences.
Around 09:00 on 13 October they confirmed unauthorized access to the system that holds member information, cut off unauthorized communication from outside, and finished containment at 13:36 the same day. The later investigation found that from about 10:00 on 9 October a third party had used the member-information retrieval API without authorization and obtained information on registered customers. On 15 October the report to Japan’s Personal Information Protection Commission was already complete. The same notice says they planned to file a damage report with the competent police station on 16 October, and that they would cooperate with the investigation. The FAQ, a separate page, says that on 16 October they consulted that police station, and that on the same day they commissioned an independent security firm to start a third-party investigation. ITmedia on 19 October says they are cooperating with the investigation. That interview does not attach a filing receipt. This page keeps “planned to file,” “consulted the police,” and “cooperating with the investigation” as three sentences.
The FAQ also answers “am I included?” Registered customers as of 13:36 on 13 October 2026 are the scope. That is a cutoff on the registration clock. It is not a promise that every person registered by then contributed every field, and it is not a page where you type an email and receive a yes or no. From 15 October they have been sending individual email, in turn, to people who may be included. People with no email address on the registration are slated for a postal letter. A mailbox that is still empty is not, by itself, proof that you sit outside the 22,519,960 records.
Items that may have left: email address, password, name, the kana reading of the name, address, gender, telephone, fax, date of birth, company name, job title, department, and system identifiers (services joined, registration date, login date, and similar). The notice says this is the full list of items that may have left, and that not every customer carried the whole table. The exact items follow what each person registered. Credit-card data is not stored in the system that was hit, and it is outside this incident. The FAQ repeats that point and adds the practical line: stopping or reissuing a card because of this incident is unnecessary.
The body of the 15 October notice does not contain “encrypted,” “hash,” or “plaintext.” The password is one item on the field list, beside the email address and the name. Whether it was encrypted is something Nikkei added on the night of the 15th, NHK on the 16th, and ITmedia on the 19th. When you read the notice, act on the field list and on “please change the other services.” A later interview does not rewrite the record count. “The key did not leave” does not delete the password from the list.
| Date | Sentence you can check |
|---|---|
| 2026-10-09 about 10:00 | Official notice: unauthorized use of the member-information retrieval API began |
| 2026-10-13 about 09:00 | Official notice: unauthorized access confirmed |
| 2026-10-13 13:36 | Official notice: containment finished; no further unauthorized access confirmed after that. FAQ: registrations up to this clock are in scope |
| 2026-10-15 | Official notice: 22,519,960 records; field list including passwords; reset asked for at login; change the same email and password on other services; report to the Personal Information Protection Commission completed; a police damage report planned for 16 October |
| 2026-10-15 19:52 | Nikkei, updated 16 October 00:19: passwords had been stored encrypted, and the program that removes the encryption was breached |
| 2026-10-16 | FAQ: consulted the competent police station, and commissioned an outside security firm. A company message quoted that day said the reset site was congested |
| 2026-10-16 17:56 | NHK headline: passwords left in a state without encryption; people were told to set them again |
| 2026-10-19 12:38 | ITmedia: the decryption key itself did not leave; passwords were confirmed to have left in a decrypted state; the company admits the attackers obtained plaintext |
The table restates nodes that already appear in those pages. It does not add the parameters of the API. It does not stretch “no unauthorized access confirmed after containment” into “the plaintext copy outside will expire on its own.”
What “left decrypted” and “the key never left” each prove
Nikkei’s story, published at 19:52 on 15 October and updated at 00:19 the next day, says the passwords had been stored encrypted, but the program that removes the encryption was breached, and the company urged people to change passwords. The sentence is Nikkei’s paraphrase. It is not a line from the notice. The visible Nikkei text also says credit-card data is managed separately and was not leaked, which sits beside the notice’s clearer line: cards are not stored in the system that was hit. Use the notice for the card field. Use Nikkei for the “stored encrypted, decryption program breached” sentence, and label it as reported speech.
NHK’s headline at 17:56 on 16 October reads: SOURCENEXT passwords left in a state without encryption; set them again. The public lead reaches “of these, the passwords had left in a state where encryption was not applied, and the company…” and then the page cuts the sentence. This page uses that headline and the half lead. It does not finish the sentence for them. Read alone, the headline contradicts “stored encrypted.” The 19 October interview is what joins the two.
ITmedia opens from a view, already circulating, that the passwords might have been stored in plaintext. The spokesperson’s answer, rendered here from ITmedia’s Japanese, says both things together. First, passwords were stored encrypted. Second, there was unauthorized access to the system (the API) that could reach the key used to decrypt those passwords. Third, the decryption key itself did not leave, but they have confirmed that the passwords left in a decrypted state. ITmedia then writes that the company admits the attackers obtained the passwords in plaintext. “Plaintext” is that article’s summary. “Decrypted state” is the spokesperson’s sentence. Both describe a password that is already readable. Another site can check that string without their key.
Reversible encryption turns a password into another string with a key. Whoever holds the key can turn it back. A one-way hash is stored so a server can check “is this the same password.” Under ordinary storage you cannot recover the password from the digest; someone who wants the original has to guess. On 19 October they explained why this system had not moved to a hash. The system was developed early. They kept the encryption method chosen at the time. As a result they had not migrated to a newer hashing method. They did not publish the year. An earlier piece on a teeth-whitening booking site followed a notice that only parenthesized “in an encrypted state,” with no algorithm and no word on whether the key left, in after a booking site marks stolen passwords as encrypted, what changing only that login still leaves on other sites. The Gyazo notice used Helpfeel’s own word, password hash, in if you upload a screenshot with a password to Gyazo, what hashes, sessions, and image text still keep. The 19 October sentences add two confirmed lines the booking-site notice did not have: the key itself did not leave, and the passwords had already left in a decrypted state. “Encrypted” on this incident can no longer be read as “people outside still cannot read it.”
“The key did not leave” proves they did not count the key file among the items already carried out. It leaves the password’s readability where the next clause puts it: already decrypted, already outside. Once that plaintext copy has left, another site’s login check does not need their key. The same sentence also leaves the future unstated. The spokesperson said the necessary security measures are already in place, and that publishing the detail could encourage a new intrusion, so the detail stays unpublished. What you can check is which of your own sites still sit on the old password. Their patch list is not a page you can audit from here.
Asked what the member-information retrieval API was for, the spokesperson declined to explain the particular implementation, again because the detail could help a new attack. ITmedia also asked about an article 山口健太 published on 17 October in a Yahoo News expert column. That piece had relayed a more specific account: that the company had told some outlet the API returned passwords because license-management and similar jobs needed them. SOURCENEXT told ITmedia that this was not an official answer they had confirmed and then given, and refused to say which part differs. The reason they gave is that they are cooperating with the investigation and will not detail the API’s purpose or how the systems connect. This page therefore leaves that column’s sentences about why the API returned passwords, and about new passwords still being encrypted the same way, off the fact list. What 19 October does record is the refusal: after you set the password again, they will not say how it is stored.
Do not read 22,519,960 as 22.52 million plaintext passwords. The notice’s number is a count of records, and the notice says not every customer carried every field. The password is among the items that may have left. What 19 October confirms is the state of the passwords that did leave: decrypted. The count and the decrypted departure both stand. On the 15th the company had already asked people to change the same email and password on other services. On the 19th they confirmed the copy was already unlocked when it left.
When you generate a replacement on this machine, MyPassGen’s random mode runs from 6 to 128 characters, with a default of 16. Below 8 characters the page warns that the result is weaker. That is the character space on this side. It is not a measurement of how SOURCENEXT stored passwords. The file box uses AES-256-GCM and stays in the browser. That choice cannot be read backwards into the algorithm in their member database. The 19 October interview did not name an algorithm. Leave the name unfilled.
Names, job titles, and mail that pretends to be them
Even if your own record happened to omit the password, the name, the kana reading, the address, the phone, the fax, the email, the date of birth, the gender, the company, the title, and the department can still be plaintext. Those fields are enough for a message that uses your name, the software you bought, the mobile number you left, and the title on the form. System identifiers also include services joined, the registration date, and the login date. An impersonation can be specific: the year you registered, a day you signed in, which product you bought. A specific detail is a field from the list. It is not proof the message came from them.
On 15 October they warned that phishing mail, SMS, calls, and postal items may pretend to be the company or a related party, and may steer you to a fake page that asks for personal information. Leave the links and the attachments unopened. Keep passwords off a page that arrived inside the message. Customers in scope receive an apology and an explanation at the registered email, in turn. People with no email on the registration are slated for post. The notice and the FAQ do not offer a box where typing an address tells you whether you are one of the 22,519,960 records. The inclusion rule they did publish is the clock: registered by 13:36 on 13 October. You apply that clock to your own registration date. This page cannot look it up for you.
The check for incoming mail is short. If the other side asks for a password, a verification code, or a card number, leave that form blank. Type sourcenext.com yourself and open /support/i/2026/1015_info/ for the notice, and /support/i/2026/1015_1/ for the FAQ. Credit cards are outside this system, and the FAQ says stopping or reissuing a card because of this incident is unnecessary. A letter that pairs “this breach” with “please bind the card again” has already stepped past the list they published. Past the list is another form to leave blank.
The FAQ also separates two products people mix up with the member password. It says Super Security and Virus Security can keep being used as before. It says passwords and other data saved in 1Password are not held by SOURCENEXT and are outside this incident, and that using 1Password is unaffected. That sentence is about the vault contents they say they never stored. It leaves untouched a password you typed into the SOURCENEXT member form and then reused on a shop. The member-password string is still the field on the 15 October list.
One FAQ answer is easy to over-read. They say membership did not collect copies of identity documents such as a driver’s license or a health-insurance card, and that a new mobile contract or a loan ordinarily requires those documents, so they consider it extremely difficult for someone to open a contract or a debt in your name from the name, email, and password alone. That is their assessment of those particular procedures. It is not a finding that the decrypted password is harmless, and it is not a finding that the mailbox is safe to leave on the old string.
If the mailbox uses the same password, change the mailbox first. Other sites send password resets to that address. The notice does not say your mail host was breached. After the mailbox has its own new password, change the shops and the other memberships that still reuse the old one. A new password only at SOURCENEXT, with the mailbox still on the old one, leaves the reset path that can pull the other accounts back.
What is still left after this login is voided
The table splits “what you already did” from “what the notice, the FAQ, and the 19 October interview named, and what can therefore still be left.” The left column is your action. The other columns stay inside sentences they have published.
| What you did | This one login | What can still remain |
|---|---|---|
| Nothing changed | The FAQ says invalidation of passwords in the leaked set is complete, and ITmedia says every customer was force-reset. When the official reset page opens, set a new password used nowhere else | A decrypted copy of the password, the email, contact details, company and title; the same password on other sites |
| You set a new password only here | That new password should stay on this site. They refused to say how it is stored | The old password can still work on other sites. Contact details and the title can still be used to write and to call |
| Before 9 October each site already had a different password | When the reset page opens, confirm this site has a new one of its own | A decrypted copy of the old password may still sit outside. It fails on the sites you already split. Contact details remain |
| You closed the account or stopped using the software | You can stop signing in here. The FAQ routes cancellation through their inquiry desk | A copy that has already left does not disappear from the other party because you cancelled |
The decrypted copy in that table rests on the 19 October sentence, “confirmed to have left in a decrypted state.” The “other sites” column rests on the 15 October request to change the same email and password on other services. Both sentences are already in public text. The voided login explains only the middle column.
Check it on this machine
Use a throwaway test password, a string you have never used on any account. Do not paste the live SOURCENEXT password, the mailbox password, or a payment password into a lookup page, a chatbot, or this article. The list on this machine does not contain the member database that was taken.
- On paper, or in a password manager, list the sites that used the same email address and password as this membership. Write the site names only. This step does not require typing the password into another web page.
- When you change those sites, type the official address yourself, or open the official app you already installed. Do not follow a link inside a “breach check” or “emergency reset” message. For SOURCENEXT, type
sourcenext.comand follow the reset guidance on the 15 October notice and the FAQ. - Open the password generator. In the developer tools, open Network and preserve the log. In random mode, take the default of 16 characters or longer, and generate one. Look at the URL, the query, and the body of each new request. The string you just generated should not be in them. Generation runs in the browser. The page opens without an account. Below 8 characters the page warns that the result is weaker. A replacement password should sit above that line.
- Write the new password into the mailbox first, then into each other service that still reuses the old one. One password per site. Do not copy this site’s new password back onto the ones you just changed. After that, copy an unrelated short sentence so the old password does not sit in the system clipboard history.
- If you want to see whether a string looks like a common weak password, open the strength check and type only the throwaway test string. The page requests
data/leaked-top10k.txt, currently about 860 lines of public weak passwords. The file opens with strings such as123456,password, andqwerty. The request body should be empty. After you type, the new requests should not contain that string. A hit means it looks like a public weak password. A miss means it is not on this list. - Back in the mailbox, look for mail pretending to be this company, Pocketalk, or a related party. Do not submit a password, a verification code, or a card number on a page that arrived inside the message. The official notice is the address you opened yourself.
Step 5 cannot tell you whether your registration is one of the 22,519,960 records. That file is a list of common weak passwords. It is not SOURCENEXT’s member database, and it is not a full-web Have I Been Pwned lookup. The password you type is not uploaded. If some page insists that you submit a live password before it will “check the breach,” you are making a second copy. The difference between a local comparison and a full-web lookup is in local leaked-password lists vs Have I Been Pwned: what each check proves.
Pasting the old password into a chatbot to ask whether it “counts as leaked” hands the plaintext to that history. Ask about the sentences in the notice and in the 19 October interview, and replace the password with a placeholder you can recognize. That boundary is already split by product switch in what to redact before you paste a password into ChatGPT or Gemini.
How to hand a new password to someone else
After you generate a new password, chat and the body of an email keep a plaintext copy if you must hand it to a family member or a colleague once. A short secret that only needs to be read once can go out as a one-time link. The key sits after the # in the URL, in the form s.html?id={id}#{key}. Browsers leave the fragment off the request line sent to the server. The server stores ciphertext. Creating the link and reading it are both public. No account is required. The full link is still the secret. A preview or a forward hands the whole string over.
Why the fragment stays off the request line is in when you send a password once, why the decryption key belongs after # in the URL. What Sent, a forward, and a lock-screen preview keep from an email body is in if you put a temporary password in an email, what Sent, forwards, and phone previews still keep. A file you need to open again belongs in the file box: AES-256-GCM on this machine, one file up to 5 GB, output as .lock or .enc. A password is not a file. Generate it, then save a different one on each site.
When the reset page will open again, give this membership a new password that has not been reused. Before that, the 15 October notice already allows the move that matters: change the other sites, and refuse to type the next password into a contact you do not trust. What 19 October adds is the reading of “stored encrypted.” The copy had already been unlocked when it left. That interview does not withdraw the request to change the other services.
FAQ
If the key never left, do the other sites still need a new password?
The 19 October sentence puts both facts on one line. The decryption key itself did not leave, and the passwords left in a decrypted state. Other sites check the password itself. A key file that stayed on their system does not expire a plaintext copy that already left. The 15 October notice already asks you to change the same email and password on other services.
This site already forced a reset. Is that the whole job?
The forced reset ITmedia reported, and the invalidation the FAQ describes, cover the old SOURCENEXT login. When the official reset page opens, set a password used only here. If the mailbox, the shops, and the other memberships are still on the old password, those login boxes still accept the string that left. They refused to describe how the new password is stored, so the new one stays off every other site as well.
A miss on the local check — does that place me outside these records?
The local list is about 860 public weak passwords, and the comparison stays in the browser. It answers whether the string looks like a weak password an attacker would try early. It does not hold this company’s member database, and it does not stand in for a full-web hash corpus. After a miss, sites that still share the old password still need a different new password each. Whether your registration is in scope is the FAQ’s clock: registered with them by 13:36 on 13 October 2026.
I have never used this software. Why does the interview still matter?
If you never registered, you sit outside the records they described. The reading of the interview still applies the next time someone says “stored encrypted, so outsiders cannot read it.” Look for a second sentence that says the passwords left in a decrypted state, and look for a request, in the same incident, to change the matching password on other services. A key that was not copied, and a password that was copied in readable form, can both be true.
Do the generator and the check require an account, and does a live password get uploaded?
No account is required. Generation and the strength check both finish in the browser. The standard you can see is Network: the word-list file may appear, and the string you typed or just generated should not appear on the request line or in the body. The words “not uploaded” on the page are confirmed by that look.