The UK Online Safety Act has been primarily discussed in the context of social media platforms and messaging apps. But the law’s reach is broader than most users realise, and in 2026, the regulator responsible for enforcing it turned its attention to cloud storage and file-sharing services.
Ofcom, the UK’s communications regulator, launched a formal investigation into whether cloud storage and file-sharing platforms had adequate measures in place to prevent the hosting and distribution of child sexual abuse material. The investigation put major storage services on notice. The regulator’s recommended technical approach involves deploying hash-matching technology to scan stored files against databases of known illegal images.
The practical implications of this policy direction are significant for anyone who stores personal files in a cloud service operating under UK jurisdiction.
What the Online Safety Act Requires
The Online Safety Act became law in 2023 and established a framework of safety obligations for digital services operating in the UK. The core obligation is that platforms must assess the risks their services could be used to facilitate illegal harm, and take proportionate measures to address those risks.
For user-to-user services — platforms where people can share content with other people — this includes proactive obligations around illegal content, including CSAM.
The critical question has always been: what does “proactive” mean in practice? Content that has already been reported can be removed without scanning everything. But proactive detection, by definition, means looking for illegal content before anyone reports it. For stored files, that means scanning the files.
Ofcom released updated guidance for file-storage and file-sharing services in late 2025, and in 2026, made clear through its investigation process that cloud storage services were within the scope of these obligations.
Hash Matching and What It Actually Does
Ofcom’s recommended technical approach is perceptual hash matching. This works by computing a hash — a numerical fingerprint — for a file, and comparing it against a database of hashes for known illegal images. The database is maintained by organisations like the Internet Watch Foundation. If a file’s hash matches one in the database, the file is flagged.
The key word is “perceptual.” Unlike cryptographic hashes, which change completely if even a single pixel is altered, perceptual hashes are designed to match images that have been slightly modified — cropped, resized, compressed, or colour-adjusted — to evade detection.
For users with legitimate personal files, this matters because:
Files must be accessible for scanning. A cloud service that encrypts files with keys only the user holds — true end-to-end encryption — cannot perform hash matching on the contents, because it cannot access the plaintext images. Compliance with scanning requirements and genuine client-side encryption are, in this technical sense, incompatible.
Scanning infrastructure creates new attack surfaces. Any system that can read and process the contents of stored files represents a potential vulnerability. A scanning system that can access plaintext images could, in a breach scenario, expose those images.
The scope of what gets hashed is not limited by user intent. A system that scans incoming files does not distinguish between users with legitimate content and those with illegal content — it processes everything.
The Encryption Conflict
The clearest casualty of mandatory content scanning is end-to-end encrypted cloud storage.
Under true client-side encryption, the service provider holds only ciphertext. It cannot read your files. It cannot scan them. It cannot comply with a technical requirement to hash and match them against CSAM databases.
Proton Drive, Tresorit, and similar services built on client-side encryption have flagged this conflict. If a regulator requires scanning, a service must either: (a) break its encryption model to enable server-side scanning; (b) shift to client-side scanning, which means running scanning code on users’ devices; or (c) withdraw the product from the regulated jurisdiction.
Option (b) — client-side scanning — is what critics have called the most dangerous path. A UK government minister stated in a parliamentary answer in 2026 that the Online Safety Act does not mandate client-side scanning of private content, and that Ofcom cannot recommend it. But the investigation into cloud storage services concerned content that is shared or accessible to others, which may be treated differently from purely private stored data.
The distinction between “private storage” and “shared content” in a cloud environment is not always clear. Files you share via a link, files in a shared folder, files accessible to others in an organisation — these are different from files stored privately and never shared. The law draws this distinction, but how it applies to consumer cloud storage products with both private and sharing functionality requires careful reading.
Which Services Are Affected
Ofcom’s investigation covered file-sharing and file-storage services operating in the UK. This includes services whose servers are located outside the UK if those services are used by UK users, under the extraterritorial reach of the Online Safety Act.
The largest consumer cloud storage services — Google Drive, OneDrive, iCloud, Dropbox — are all within scope as services that allow file sharing. Services with a UK user base above certain thresholds have additional obligations, including designated Category 1 or 2 service status.
UK-based services or services primarily serving UK users face the most direct regulatory exposure. Services based entirely outside the UK with no UK establishment have a more complex legal relationship with the act, though Ofcom has pursued enforcement actions against non-UK services in other contexts.
The Privacy Argument Against Mandatory Scanning
Privacy organisations including Open Rights Group and the Electronic Frontier Foundation have been consistent in their opposition to mandatory content scanning requirements in cloud storage. Their arguments:
Mass scanning is disproportionate. The vast majority of stored files are entirely lawful. Scanning every file to find the small number of illegal ones treats all users as suspects rather than the bad actors who are the actual target.
Scanning infrastructure cannot be limited in scope. A system capable of scanning files for CSAM hashes is a system capable of scanning for other content. Once the infrastructure exists, the question of what it scans is a policy decision, not a technical constraint.
Accuracy rates matter at scale. Even very accurate hash-matching systems produce false positives when applied to billions of files. Family photos that superficially resemble flagged images, benign images of children in bathing situations, innocent photographs with similar perceptual characteristics — these can generate flags that then require human review. Human review of flagged content means someone employed by the service sees your photos.
What This Means for UK Cloud Storage Users
If you store personal files in a cloud service and you are in the UK, the practical picture is:
Services that use server-side encryption will likely comply. A service that holds your encryption keys — most major consumer cloud storage products — can implement hash matching without changing their architecture. Your files are already accessible to their systems.
Services with true client-side encryption face a choice. Some will comply by modifying their architecture. Some may limit certain features for UK users. A small number may withdraw from the UK market or restrict UK sign-ups.
The nature of “shared” content matters. Files you keep entirely private and never share through any mechanism may fall under different rules than files you share via link or collaborative folder.
The law is still being interpreted. Ofcom’s codes of practice and the courts will continue to shape what exactly is required. The 2026 investigation is early in that process.
The Broader Pattern
The UK is not alone in pushing for scanning of cloud-stored content. The EU’s proposed Chat Control regulation has sought to extend similar requirements to private communications, though it has faced substantial opposition and legal challenges. India’s data protection framework includes provisions that could similarly require service providers to access stored data for law enforcement purposes.
The direction of travel in multiple jurisdictions is toward greater regulatory access to content stored in cloud services. The argument for such requirements — protecting children from exploitation — is not frivolous. The question is whether mandatory scanning of all stored content is proportionate to that goal, and what the privacy costs of that choice are for the billions of people storing personal files in the cloud.
For users who have not thought carefully about where their personal files live and how they are protected, the answer to these questions is now more consequential than it was a few years ago. A service that stores your personal photos and can be compelled to scan them is a different kind of service from one that cannot.
Understanding which category your cloud storage provider falls into is a reasonable thing to know.