A routine migration, an unreviewed consequence
The root cause traces to February 4, 2025, when DigiCert moved an internal business archive from a network segment restricted to VPN access into a different availability zone during an infrastructure migration. That move made the archive reachable from the public internet, a consequence that apparently went unreviewed for well over a year. Compounding the exposure, a self registration feature designed for an unrelated application was reachable from the same system, and it did not require the multi factor authentication that DigiCert mandates for staff accounts.
A permissions flaw then granted any self registered account default read only access to the archived customer files, no approval step, no access request, just an open door behind an unlocked front gate. This is the kind of failure that single point security reviews are built to catch: a migration changes network placement, an unrelated feature happens to be reachable from the new location, and a permissions default nobody revisited compounds the exposure. Each piece in isolation might have been low risk. Stacked together, they produced over a year of open access to sensitive archives.
What was actually exposed
The majority of affected records held standard business contact information: names, corporate email addresses, phone numbers, and physical business addresses, the kind of data that shows up in any B2B vendor relationship and carries real but modest risk. A smaller subset of customer files is more serious. DigiCert, as a certificate authority, requires identity verification documents during its vetting process for certain certificate types, and some of the exposed archive contained driver's licenses, passports, and utility bills submitted for that purpose.
Cloud architect Fabian Bader captured the gap between DigiCert's characterization and the reality in a public post: thanks for a data leak that was limited to only full personal identifiable information. That reaction is the predictable result when a company describes identity document exposure using the same measured language it would use for a marketing list leak. For a certificate authority specifically, whose entire business model rests on being trusted to verify identity correctly, exposing the documents used for that verification is a direct hit to the core value proposition, not a peripheral incident.
The detection and disclosure timeline is the real story
DigiCert noticed non company accounts on the system on June 18, 2026, and responded by adding a web application firewall, a mitigation rather than a fix. It received an external report through its managed security research program on June 21. The confirmed window of unauthorized read only access ran from June 27 to July 2, when the authorization fix finally deployed. DigiCert did not determine the incident was reportable until October 19, three and a half months after the access window it confirmed, and sixteen months after the original exposure began.
Each gap in that chain represents a decision point where the incident could have been triaged faster. Noticing anomalous accounts and deploying a firewall as the first response, rather than immediately auditing what those accounts could access, is the kind of instinct that treats a security signal as an access control problem to patch rather than a data exposure to investigate. The four month gap between confirming unauthorized access and determining reportability is harder to explain and suggests the investigation itself, not just the original misconfiguration, moved slower than the sensitivity of the data warranted.
Why this matters more for a certificate authority than almost anyone else
DigiCert's entire function in the internet's trust infrastructure depends on correctly verifying who it is issuing certificates to. The identity documents in that exposed archive exist specifically to support that verification process, which means this breach touches the one asset category a certificate authority cannot afford to mishandle. DigiCert's statement that certificate authority services, login credentials, and issued certificates were unaffected is true and also beside the point: the documents proving identity are themselves the product of the verification process customers pay for.
Enterprises that rely on DigiCert or any certificate authority for code signing, TLS issuance, or document signing should treat this as a prompt to ask their vendor directly how identity documents are stored, for how long, and under what access controls, rather than assuming a reputable CA has that process locked down by default. The sixteen month exposure window here shows that assumption does not hold even at the top of the market.
The build versus buy question this raises for internal tooling
The proximate cause was an internal business archive, internal tooling that DigiCert built or configured itself, not a third party vendor. That matters for how enterprises read this incident: the usual deflection of blaming an external vendor does not apply here. This was DigiCert's own infrastructure migration, its own self registration feature, and its own permissions default that stacked into sixteen months of exposure. A company that spends heavily on external vendor audits while leaving its own internal tooling unreviewed has simply moved the risk rather than reduced it.
The lesson for any enterprise running internal archives or admin tooling is that infrastructure migrations need an explicit security review step that checks network reachability and authentication requirements against the sensitivity of what is being moved, every time, regardless of how routine the migration seems. A checklist that would have caught this costs an afternoon per migration. The incident it would have prevented cost DigiCert sixteen months of exposure and a public relations problem that goes to the heart of what the company sells. Weigh that afternoon against that cost the next time a migration ticket gets marked routine and waved through without a security sign off, because routine is precisely the label this migration carried too.



