When several businesses use the same online service, their information should still stay separate. A recent Cloudflare report explains how a mistake in one part of its storage system could have weakened that separation for some customers using Cloudflare Containers and Cloudflare Sandboxes.
Cloudflare says a security researcher reported the issue on September 4, 2026. The company has since fixed it across the affected systems. It says customers do not need to change anything, and it found no evidence that someone had deliberately used the flaw to steal customer information.
How could old information appear?
Imagine a shared office with lockable desks. When one person leaves, their desk should be cleared before it is assigned to someone else. The reported problem was a little like a desk that was reused before every last scrap had been removed.
Cloudflare’s online service gives each customer a separate working space on a larger computer. Those spaces use storage to hold information while they are running. When a space is removed, its storage can later be reused.
In this case, a small new piece of information could be written into part of a reused storage area without first clearing all of it. The unwritten part could still contain scraps left by an earlier customer’s space. A researcher showed that, in some circumstances, a customer using a paid Cloudflare account could see some of those leftover pieces if their new space happened to be placed on the same underlying computer.
That did not mean a person could choose whose information to look for. The placement was automatic. The researcher could not pick a particular customer, and there was no guarantee that any old information would be there. Cloudflare says the issue affected Containers and Sandboxes built on that service.
What did the researchers find?
The researchers tested the issue in several production placements and found signs of information from other customers’ past storage. Their report describes material such as folder records and parts of database files. That shows why a storage mistake can matter even when customers are meant to be kept apart.
There are important limits to what the findings show. Cloudflare says it found no evidence of malicious use in the historical records it could check. The activity it could connect to the method came from the researchers and Cloudflare staff carrying out authorized tests. The company also says the researchers’ submitted materials did not contain customer names, account details, passwords, or the recovered information itself.
That is not the same as saying the design was harmless. If leftover information can be seen by someone who should not see it, the system needs to be corrected. It is also important to be careful with the wording: Cloudflare reported no evidence that customer information had been compromised, not proof that every possible exposure can be ruled out.
What changed—and what should customers do?
Cloudflare says it changed how new storage is prepared, then replaced older stored copies that might still have been affected. It has completed those steps across the service. Customers do not need to make a setting change or take another action for this issue.
The wider lesson is simple: information should be cleared before storage is reused, and companies should check both new and older stored copies when they find a problem. A clear public explanation also matters. Customers need to know what was affected, what has been fixed, what evidence the company reviewed, and whether they need to act.
Cloudflare credited Oren Yomtov and the Accomplish security research team for reporting the issue through its bug bounty program. Their careful testing helped the company confirm the problem and respond.
At Evryday Studio, we pay special attention to this: keeping business information protected and explaining technology risks in plain language.
