Our full report on the October incident is below. It’s also on X: FOUNDATION on X: "The October security incident: what happened and what comes next" / X
The October security incident: what happened and what comes next
Late on Sunday, October 25, Foundation customers began receiving emails claiming that their Passport had been recalled. The campaign involved more than 23,000 phishing email attempts targeting 21,463 email addresses. The messages carried our branding and my name, came from real Foundation email addresses, and, in some cases, included the recipient’s purchase details. They directed people to websites asking for their recovery phrases.
Seven minutes after the first internal report, our team was on a call investigating the emails. We began revoking credentials and removing the attacker’s fraudulent web address, then worked through the night to stop the remaining messages, isolate the compromised application, and warn customers.
We’re sorry for the concern and harm this has caused. Our earlier emails shared what we knew as the investigation progressed; this report brings that information together and explains the changes now underway. We also want to address a question that keeps coming up: how did this happen when we self-hosted our services, kept internal systems behind a VPN, and protected employee accounts with two-factor authentication?
Funds held on Passport are safe. The phishing emails and calls are attempts to steal them by getting you to reveal your recovery phrase or send funds to the attacker.
Our response that night
Our first internal alert arrived at 11:05 p.m. Eastern on October 25. By 11:12 p.m., we were on a call reviewing the emails and investigating the unauthorized subdomain under foundation.xyz. Within about ten minutes of the alert, we were removing the malicious DNS record and revoking Cloudflare credentials.
By about 11:30 p.m., we had revoked the keys the attacker used to send through Mailgun, our email provider. Shortly afterward, we confirmed that the fraudulent Foundation subdomain had been removed. We stopped the remaining emails already queued for delivery around 12:40 a.m.; by 12:45 a.m., we had traced the attack back to our Cal.com booking application and isolated it.
The work continued through the early morning. We removed affected Shopify integrations, took the archived WooCommerce service offline, revoked Google email app passwords, and replaced database and storage credentials. We saved logs and copies of affected data for the investigation, restricted administrative access, and paused releases while we reviewed the systems involved.
Alongside that work, we published a warning on X, added a notice to our website, and sent the first customer warning email. We then compared the phishing recipient list with our mailing list so we could contact people who had not been covered by that first notice.
All times above are Eastern. In UTC, this was October 26 – the date used in our earlier updates.
How the attacker got in
The attack began in our self-hosted Cal.com installation, which we used for booking appointments. Our investigation points to a code-execution vulnerability – malicious input submitted through the booking application was designed to run commands on its server. The earliest confirmed unauthorized activity dates to October 21, several days before the phishing emails were sent.
From Cal.com, the attacker exploited a separate vulnerability in Bank-Vaults, the software connecting our applications to HashiCorp Vault. Vault held the passwords and keys those applications used to access other services. The attacker tricked the vulnerable component into handing over another service’s login credential, then used it to enter Vault and retrieve credentials for additional applications.
Those credentials allowed the attacker to read customer records, send emails through our providers, and change our domain settings. In Shopify, for example, the attacker used stolen credentials belonging to applications connected to our store.
The compromise also reached Google Workspace mailboxes used for support and hiring. Historical correspondence in those accounts gave the attacker access to email addresses beyond our customer and mailing lists. That is why people who had only contacted support or applied for a job also received phishing emails.
Why the VPN was not enough
Like many in the Bitcoin community, we preferred to run our own software. We self-hosted our services, used WireGuard to access internal systems, and enabled two-factor authentication for employee accounts. Even our appointment-booking website ran on our own infrastructure.
After moving to Shopify, we preserved the old WooCommerce store as an internal archive so our support team could look up earlier orders. Staff accessed it through the company VPN, and older orders appeared redacted in the dashboard. It was no longer available as a public site.
The archive was still a running application connected to a database. Cal.com ran within the same hosted environment and remained publicly accessible so customers could book appointments. Once the attacker could run commands through Cal.com, they could reach the vulnerable internal component and obtain credentials for other services.
Our WireGuard VPN restricted access from outside our network, but the attacker was making requests from an application already running inside it. The stolen credentials were also designed for software to use automatically, so the attacker did not need an employee to approve a two-factor login. (No employee accounts or devices were compromised.)
Over the past several days, people have suggested that we put everything behind a VPN. We understand the reasoning because that was already how we protected these internal services. What happened here has changed how we think about the access each application has to the rest of our infrastructure.
What information was affected
The attacker attempted to send more than 23,000 phishing emails to 21,463 distinct email addresses. Some recipients received multiple messages; other messages failed or were stopped before delivery.
The information affected included names, email addresses, mailing addresses, phone numbers, order details, archived support conversations, and the payment information described below.
Our earlier updates reported approximately 5,600 mailing addresses and 5,200 phone numbers as potentially exposed. These were estimates from our initial review. We have since identified additional customer and payment records and are still determining how much they overlap with the information already counted.
More than 5,000 addresses targeted by the phishing emails were absent from our store records. Compromised email accounts supplied additional addresses, and customer reports also point to information obtained from outside Foundation. We explain both below.
Why older customer information was still present
Before moving to Shopify in May 2026, our store ran on WooCommerce. Our policy was to redact customer details 60 days after shipping, and we had configured automatic redaction for that purpose. When we viewed older orders in the store dashboard, the customer details were no longer visible and the orders appeared redacted.
During our investigation, we discovered that WooCommerce had automatically migrated orders into newer database tables while leaving the original copies in the old tables. We found that redaction had updated the newer records, but older copies still contained customer names, addresses, and phone numbers. Our dashboard only displayed the newer records – which explained why those orders had appeared redacted when we viewed them.
We found information in other places too. Some order and refund statuses were not covered by the retention process, and plugins stored phone numbers in separate fields that survived after the main order fields were cleared. Abandoned carts and contact forms also retained details entered by people who never completed a purchase.
When we moved to Shopify, we updated our privacy policy because Shopify did not offer a built-in way to automatically remove customer details from orders after a set period. We were building our own tool to do that, but it was not yet running when the incident occurred. Customer details from those newer orders had therefore remained in the store.
We’re continuing work on the automatic deletion tool for Shopify. It has passed initial testing, and we are completing the security and deployment checks before enabling it on customer records.
We have taken the affected legacy services offline and kept evidence separately for the investigation. We intend to permanently delete the retained WooCommerce evidence once the investigation and preservation requirements allow. Our retention work includes checking the database records and connected systems themselves.
We also plan to share these findings with the relevant software providers.
What payment information was affected
Our investigation also identified access to payment records containing partial card information: the cardholder name, last four digits, and expiration month and year. Full card numbers were handled by the payment processors and were never available to Foundation.
For our current store, which moved to Shopify in May 2026, the attacker retrieved transaction records using stolen app credentials. Our previous WooCommerce store used Stripe to process card payments; credentials associated with that older integration were used to access historical customer and payment records in Stripe.
Stripe keeps its payment history separately from our store. Our review found a payment from 2020 that still contained partial card details even after the related customer record in Stripe had been deleted.
Stripe says it generally retains personal information for five or more years in most jurisdictions, measured from the end of the business relationship or the last transaction, whichever comes later. Its reasons include legal requirements and fraud prevention. Stripe’s retention policy.
Customers have reported callers quoting their card’s last four digits and expiration date alongside their purchase details.
How the phishing emails and websites worked
The emails claimed that a supplier had modified a component in Passport and that customer funds were being stolen. They urged people to act immediately, presenting the websites as an emergency recall process. Some included a name, mailing address, or phone number; another version used our own retention policy as part of the story, saying we had deleted the customer’s address and needed them to enter it again.
The attacker sent the messages through our Mailgun and Google Workspace services using stolen credentials. They therefore came from real Foundation addresses, and the Mailgun messages passed the usual email-authentication checks. To a recipient, the sender address, design, signature, and personal details could all look familiar – even though the message itself was fraudulent.
The links took visitors through redirects to phishing pages. The attacker had also used stolen DNS credentials to create incident[.]foundation[.]xyz, a subdomain used by the phishing operation. Other pages appeared on lookalike domains such as foundation[.]support and foundationwallet[.]xyz.
The websites checked visitors’ browsers before deciding what to display. We found code that sent a browser fingerprint to the attacker’s server; a visitor could then receive the fake recall page or be redirected to our real website. Someone checking a reported link could therefore land on our store while a customer saw a page asking for their recovery phrase. The public scans we reviewed were redirected to the real Foundation site.
The software also allowed operators to view visitors, respond through live chat, accept callback requests, and receive notifications when a recovery phrase was submitted. A customer who had questions or trouble following the fake recall process could be guided through it by someone pretending to provide support.
Why people outside our mailing list were targeted
The attacker’s recipient list was larger than the number of email addresses stolen from Foundation. Some recipients told us they had never purchased a Foundation product, subscribed to our newsletter, or otherwise interacted with us, yet had received a personalized email pretending to come from Foundation.
Other customers showed us details that did not match their Foundation orders. These included old addresses and addresses they said they had never given us. One recipient said the address had only been supplied to another hardware-wallet company.
We believe the attacker combined information stolen from Foundation with information from other sources. Matching an email address across different datasets could connect a Foundation purchase with a phone number or an old mailing address that the customer never gave us.
The calls and follow-up scams
Customers have reported calls from scammers claiming to be Foundation support, sometimes passing them to a second caller introduced as a manager. The callers quote names, addresses, order numbers, purchase dates, and sometimes the last four digits and expiration date of a payment card. They may provide a reference number and ask the customer to repeat it back, much like a routine support call.
The story is usually that a Passport has a manufacturing fault, is part of a recall, or needs to be replaced as a precaution. The caller offers a replacement or refund, then asks the customer to “verify” their recovery phrase, visit a fake website, or move funds into a supposed secure vault. Some say they cannot process the replacement until the customer completes that step; others start with an automated “security notice” asking the recipient to press 1.
The callers use those purchase details to make the conversation sound like a normal support call, then ask the customer to reveal a recovery phrase or move funds.
Our November 2 warning listed reported caller aliases, phone numbers, and domains, but these can change. Reports have included foundation[.]holdings and emails from newinblue[.]com and auguste[.]io. Callers can also fake caller ID.
We have also seen scammers posing as the International Trade Centre tell victims to buy our Concierge Pro service and install remote-access software such as AnyDesk. Directing someone to purchase a real service on our website appears to be part of making an investment scam look legitimate. Foundation does not offer trading or investment services.
We refunded the identified Concierge Pro purchases associated with that scam and warned the purchasers.
The work since that night
After tracing how the attacker obtained the service credential, we disabled the vulnerable Bank-Vaults component and removed it from our configuration. Follow-up checks confirmed that the retired components were no longer present. We continued replacing credentials, restricting permissions, and reviewing connected applications individually.
We replaced database and storage credentials and, for several services, verified that the old credentials no longer worked. We also revoked legacy payment integrations and asked the affected providers to investigate and preserve their records. Our review extends to credentials saved inside older applications, since shutting down an application does not automatically disconnect its access to outside services.
Our support team has continued responding to customers and reporting fraudulent sites and senders. We identified 8,953 addresses on the phishing list that were not covered by our first warning and used that list for another mailing. Some messages failed or were suppressed by the email provider, so we also continued posting updates on X and our community forum. On November 2, we sent a further warning about the phone scams customers were reporting.
Why we are changing our approach to self-hosting
For years, we believed that running our own services gave us more control over our software and data. We applied that belief throughout Foundation, including to applications that most companies would simply subscribe to.
This incident has changed our view. We are moving from self-hosting as much as possible to self-hosting as little as possible.
Running those services meant maintaining much more than the applications themselves. It meant servers, databases, deployment tools, passwords, and permissions connecting everything together. We want to reduce how much of that infrastructure we operate and how much access any one application can have.
We are now reviewing every self-hosted service and moving suitable applications to managed infrastructure, including Cloudflare Workers. Several replacements are already running in test environments; we are validating them before moving production traffic.
With Workers, we can run application code without maintaining a dedicated server or cluster. We can also connect an application to the specific database it needs without giving it credentials for unrelated services. Our team still needs to secure the code and its dependencies, while the provider operates more of the underlying infrastructure.
We are reviewing deployment access and the permissions each service receives, along with limits and alerts for activities such as sending email. We are also retiring services we no longer need and testing how their replacements recover from an outage.
This is a major overhaul, and it is underway. We will move production services in stages as testing is completed.
What customers should do
There is no recall or replacement program associated with these messages. Foundation will never ask for your recovery phrase.
If you receive an unexpected call claiming to be Foundation, hang up. Never share your recovery phrase, PIN, or passphrase, and never enter your recovery phrase into a website. Do not move funds or install software at a caller’s direction, even if they know your purchase details.
If a device you did not order arrives, do not turn it on or connect it, and never enter your recovery phrase on it. Photograph the packaging and contact us. As of November 2, we had no reports of physical replacement devices arriving through this scam.
If you have already shared your recovery phrase, treat that wallet as compromised. Move any remaining funds to a new wallet with a new recovery phrase, set up on a device you trust, and contact us through our official support channel. Do not use a wallet or replacement phrase supplied by the caller.
To contact us, type foundation.xyz into your browser yourself and use the support options there, or email hello@foundation.xyz.
Thank you to everyone who reported suspicious messages and helped us respond. We’ll continue sharing what we learn and updating you on the infrastructure work.
Zach Herbert, CEO