In May 2026, researchers at Mysterium VPN published findings from a scan of cloud storage infrastructure across Amazon Web Services, Google Cloud, Microsoft Azure, DigitalOcean, and Alibaba. Using a March 2026 snapshot, they identified approximately 535,480 publicly accessible cloud storage buckets — no authentication required — containing around 19.6 billion files.
Five months later, researchers at Ransomnews published an updated count from a scan conducted on October 7. They found 484,142 publicly listable cloud storage buckets containing 18.05 billion files. Amazon S3 hosts roughly 72% of the exposed buckets. About 67% of the exposed configuration files — the .env files that often contain application secrets, API keys, and database credentials — had been last modified in the current year.
Neither research team accessed or downloaded the files. They counted what was publicly listable. The data existed in states where any person with internet access could read or download it.
This is not a new problem. Misconfigured cloud storage has appeared in high-profile breaches for years. What the 2026 studies establish is the scale of what remains exposed after years of awareness campaigns and provider-level controls: billions of files, hundreds of thousands of buckets, available right now.
What’s Actually in These Buckets
“19.6 billion files” is an abstract number. The concrete question is what kinds of files end up exposed.
Security researchers who investigate specific open buckets consistently find the same categories:
Personal photos and videos. When app developers store user-uploaded media, they create a storage bucket for it. If that bucket is misconfigured — set to public when it should be private — every photo a user uploads to that app is accessible to anyone. The AI photo app exposures from February 2026, which affected over 8 million media files, illustrate exactly this mechanism.
Government ID documents. KYC verification apps, age-verification services, and financial platforms often use third-party storage for the ID photos users upload. These are among the most sensitive documents a person can have exposed — they contain name, date of birth, address, and ID number in a photo format that’s immediately usable for identity fraud.
Medical records and insurance documents. Healthcare-adjacent apps and patient portals have appeared repeatedly in open-bucket research. Medical records are regulated under HIPAA in the United States, but the regulation applies to covered entities and their business associates — not necessarily to every app that collects health-adjacent data.
Financial documents. Bank statements, tax documents, and financial records submitted for loan applications or KYC processes have appeared in open buckets associated with fintech apps. A financial document combines personal identity, address, account numbers, and behavioral data in a single file.
Backup files. Developers and small companies that use cloud storage for backups sometimes misconfigure the backup destination. Database dumps, configuration archives, and code repositories with embedded secrets appear in open-bucket research at significant volume.
App configuration files. The .env files Ransomnews specifically flagged contain credentials that let attackers access the underlying application — not just the exposed storage bucket, but the entire service. An exposed .env file can hand an attacker account-level access to a cloud infrastructure they could use to access or modify every other file in the system.
How Misconfiguration Happens at Scale
Understanding how 19.6 billion files end up publicly accessible requires understanding how cloud storage configuration errors happen — and why provider controls haven’t fully solved the problem.
Cloud storage services work through a system of buckets (in AWS’s terminology), containers (Azure), or blobs (Google Cloud), each of which has an access policy that determines who can read or write its contents. The access policy can be set to private (only specified accounts), to public read (anyone can read), or to various intermediate configurations.
Testing defaults that never get changed. A common pattern: a developer creates a storage bucket in a testing environment, sets it to public access for ease of testing, and later deploys the same configuration to production without remembering to update the access policy. Public access that was acceptable for a test environment becomes public access for live user data.
Inherited configurations. When a team inherits code or infrastructure from another developer, they may not review or understand every setting. A public access configuration buried in an infrastructure-as-code file may not be obvious to someone who didn’t write it.
Deliberate temporary access that persists. Sometimes public access is intentional for a specific purpose — serving static assets for a website, or allowing temporary access for a specific workflow. The intent is temporary; the configuration persists.
Multiple bucket, multiple team issues. At larger organizations, different teams may manage different buckets with different levels of security rigor. A misconfigured bucket created by one team may not surface in another team’s security audit.
Why Provider Controls Haven’t Solved This
AWS introduced its S3 Block Public Access feature in 2018. When enabled at the account or organization level, it prevents any bucket from being configured for public access, even if someone tries to set one that way. Google Cloud and Microsoft Azure have equivalent controls.
These features work well when they’re applied. The Mysterium and Ransomnews 2026 findings indicate they are not universally applied.
Part of the reason is that the controls must be deliberately enabled — they’re not retroactive on existing buckets, and they may not be part of every organization’s default cloud setup. A company that created a cloud account before these controls existed, or that never applied them during a migration, may have buckets that predate the account-level protection.
Part of the reason is the long tail of developers. Large companies with dedicated security teams generally apply these controls. The buckets in the 2026 research likely disproportionately belong to smaller organizations, individual developers, and apps that never reached the scale that would trigger formal security review.
Are Your Files at Risk?
If you use cloud storage services directly — iCloud Drive, Google Drive, Dropbox, OneDrive — your personal files are not at risk from this particular mechanism. These consumer cloud services don’t let users set their own bucket access policies; the storage infrastructure is managed by the provider, and the access controls are handled at the account and sharing layer.
The risk is indirect: when you upload your personal files to apps that store them in cloud storage they manage, you’re relying on those apps’ backend infrastructure to be correctly configured. That’s the category the 2026 research and the February photo app exposures document.
The relevant question for each app you use to store photos or personal documents is: where does the app actually store the files you upload? This is almost never disclosed transparently, and most app privacy policies describe storage practices in terms general enough to be nearly uninformative.
Signals that suggest stronger backend practices:
- The developer is a larger, security-mature organization with a formal security program
- The app has undergone third-party security audits (occasionally disclosed in privacy policies or dedicated security pages)
- The privacy policy is specific about data retention and storage, rather than vague about “secure storage”
- The app has a security disclosure process, sometimes called a responsible disclosure or vulnerability reporting program
Signals that suggest higher risk:
- A small or new developer with no disclosed security practices
- Privacy policy boilerplate with no specific storage or retention information
- Broad permission requests that exceed what the app’s function requires
- No disclosed third-party audit or security review
What Cloud Storage Providers Are Doing
The 2026 research reached providers as well as researchers. AWS has continued to expand its S3 security defaults and in recent years began notifying account holders when new publicly accessible buckets are created. Google Cloud has implemented similar notifications in its Security Command Center.
These are improvements. They don’t retroactively secure misconfigured buckets that were created before the notification systems existed, and they don’t cover every cloud provider globally.
The researchers’ recommendation in both studies was consistent: enable provider-level public access controls at the account or organization level, apply them retroactively to existing buckets, and treat any bucket that needs public access for a specific function as an exception that requires explicit documentation.
For the people whose files are already in those 19.6 billion exposed items: the studies don’t identify affected individuals, and there’s no way for an individual to check whether a specific file from a specific app is in an open bucket. The practical response is the same as for any unconfirmed data exposure — monitor for identity fraud signals, use unique email addresses for app accounts where possible, and treat any app that holds sensitive personal files as a potential exposure vector.
The Underlying Point About Where Personal Files Live
The scale of the 2026 findings — nearly 20 billion files in open storage — is an illustration of a broader point about the difference between data you store directly and data that gets stored as a consequence of using apps.
When you upload a file to a service you’ve explicitly chosen for personal storage, you can evaluate that service’s security practices, review their audit records, and make an informed choice about trust. When an app stores files in its own backend as part of its operation, you’re trusting infrastructure you have no visibility into.
daftei stores your files with AES-256 encryption at rest and TLS 1.3 in transit. Files are stored in infrastructure that daftei controls and is responsible for, not passed to third-party backends whose configuration you can’t review. The 5 GB free tier and Pro plan both operate under the same encryption and privacy standards.
The distinction matters not because other services are necessarily less secure, but because the 2026 research documented what happens when backend infrastructure isn’t managed with the same care as the consumer-facing product. Files in the category of “uploaded to an app” carry different risk than files in the category of “stored in a service you explicitly chose for that purpose.”