If you’ve spent any time comparing cloud storage services recently, you’ve encountered the word “encrypted” used to describe almost every option on the market. Proton Drive is encrypted. Dropbox is encrypted. iCloud is encrypted. Google Drive is encrypted.
These statements are all technically true. They’re also not equivalent. The encryption provided by these services protects you from meaningfully different threats, and in some cases provides essentially no protection against the threat you’re most likely to care about: the service provider itself accessing your data.
Understanding the difference between zero-knowledge encryption and server-side encryption is not a technicality for security researchers. It’s a foundational question about what “private storage” actually means.
Server-Side Encryption: What It Does and Doesn’t Protect
The majority of cloud storage services that describe themselves as encrypted are using server-side encryption — sometimes also called “encryption at rest.”
Here’s how it works: when you upload a file, it travels over an encrypted connection (HTTPS/TLS) to the service’s servers. Once it arrives, the provider encrypts it using a key that the provider stores and manages. Your file sits on disk in encrypted form.
This protects you from one specific threat: someone gaining raw access to the storage hardware. If a hard drive is physically stolen from a data centre, or if an attacker accesses raw storage through an infrastructure breach without proper access controls, encrypted data on that drive is not readable. That’s the specific threat server-side encryption is designed to address.
What it does not protect you from:
The provider can decrypt your files whenever it chooses to. The provider holds the keys. This means:
- Legal requests. A government subpoena, court order, or national security letter can compel the provider to hand over your decrypted files — or simply hand over the keys. The provider cannot refuse by claiming it doesn’t have access. It does have access.
- Insider threats. An employee with sufficient access rights can, in principle, access decrypted files. Most providers have internal controls and auditing to deter this, but the technical capability exists.
- Policy changes. If a provider decides to scan your files for content (for AI training, for advertising, for safety review), server-side encryption doesn’t prevent this. The provider can decrypt and read files as a matter of normal operations.
- Provider-level breaches. A sophisticated breach that compromises the provider’s key management system, not just the storage layer, could expose decrypted data.
Server-side encryption is a real security measure. It protects against real threats. It should not be confused with privacy from the provider itself.
Zero-Knowledge Encryption: What Changes
Zero-knowledge encryption — also called client-side encryption or end-to-end encryption, depending on the context — works differently at a fundamental level.
With zero-knowledge encryption, your device encrypts the file before it leaves your control. The ciphertext — the scrambled, unreadable version of your file — is what gets uploaded to the provider’s servers. The encryption keys are derived from your passphrase or stored on your device only. The provider’s servers never receive the plaintext version of your file, and they never receive your keys.
The result is that the provider genuinely cannot read your files. Not as a policy commitment, but as a technical fact. There’s nothing to hand over to a government request beyond encrypted data that is unreadable without the keys. There’s nothing for an insider to access. There’s nothing for an AI system to scan. The data is mathematically opaque to anyone who doesn’t hold your keys.
This is where the term “zero-knowledge” comes from: the provider has zero knowledge of the contents of what you’ve stored.
Services that offer genuine zero-knowledge encryption include Proton Drive, Tresorit, Internxt, and Sync.com — all of which encrypt files client-side by default. MEGA offers client-side encryption. Some services, like pCloud, offer it selectively for a designated encrypted folder rather than the entire storage account.
The Real Difference, Practically Speaking
The distinction becomes concrete in three scenarios:
A government subpoena
With server-side encryption, the provider receives a subpoena and hands over your files in readable form. They may notify you (some services have transparency report commitments), but they technically can comply. This is not a hypothetical — cloud storage services receive tens of thousands of government requests annually, and most comply with legally valid demands.
With zero-knowledge encryption, the provider receives a subpoena and has nothing useful to hand over. They can provide encrypted blobs. Compelling the provider doesn’t get a government access to your files because the provider doesn’t have the keys. This shifts the legal demand to you — which may or may not be better, depending on your circumstances — but it means the provider cannot hand over what they don’t have.
A data breach
With server-side encryption, a breach that reaches the key management layer — not just the storage layer — could expose plaintext data. With zero-knowledge encryption, even a complete server-side breach exposes only encrypted data, because the keys were never on the server.
The provider scanning your files
With server-side encryption, the provider can run their systems — for moderation, for AI training, for feature development — against the decrypted content of your files. With zero-knowledge encryption, this is technically impossible without breaking the encryption first.
What Zero-Knowledge Encryption Costs You
This is where most comparisons stop being honest. Zero-knowledge encryption is not obviously better in every dimension. It comes with real trade-offs.
No AI-powered features
If the service can’t read your files, it can’t search their contents, auto-tag photos, generate smart albums, or run the kind of AI-powered organisation features that services like Google Photos have made familiar. Zero-knowledge encryption and server-side AI assistance are technically incompatible unless the AI processing happens entirely on your device.
This is not a marginal limitation. For many users, the ability to search a photo library by “sunset in Portugal” or find a document by its contents is the primary value proposition of cloud storage. That functionality requires the server to see the content.
Slower uploads and syncing
Client-side encryption adds computational overhead on the upload path. Your device is doing cryptographic work before bytes are sent. For large photo or video archives, this can be measurably slower than uploading directly to a server-side-encrypted service.
No server-side deduplication
Services like Dropbox use a technique called content-based deduplication — if you upload a file that’s already in the system under another account, they can detect this and avoid storing a second copy, which saves storage costs and makes uploads appear instant. Zero-knowledge encryption breaks this: since all files are opaque on the server, the provider can’t identify duplicates.
Recovery is harder
If you forget your passphrase and there’s no recovery mechanism, zero-knowledge encryption means the provider genuinely can’t help you. Your files are gone. Server-side encrypted services typically let you recover your account through email verification, because they hold the keys and can reset your access.
Sharing is more complex
Sharing a zero-knowledge encrypted file requires key management — either sharing the key directly (which has its own security implications) or using a complex key-sharing protocol. Server-side encrypted services can share files simply because the provider mediates the access.
How to Find Out Which Type a Service Uses
The marketing doesn’t always make this clear. Here’s how to find out:
Look for “client-side” or “end-to-end” encryption. Server-side services typically say “encrypted at rest” or “AES-256 encryption.” Client-side services will specify that encryption happens on your device, that they have no access to your keys, or that they use “zero-knowledge architecture.”
Read the privacy policy section on data access. A service using server-side encryption will typically acknowledge that it can access your data in limited circumstances — for safety review, legal compliance, or abuse prevention. A genuine zero-knowledge service will state that access is technically impossible.
Ask whether the service can recover your account if you lose your password. If yes, they hold your keys — they’re not zero-knowledge. If the answer is “no, you should save your recovery phrase,” the service is likely genuinely zero-knowledge.
Check whether AI-powered search and organisation features exist. If a service offers in-library content search or AI-powered photo tagging, server-side access to files is required. These features and genuine zero-knowledge encryption are mutually exclusive.
The Middle Options
Not every storage service is purely one or the other. Some approaches sit between the two models:
Selective client-side encryption. pCloud’s Crypto Folder and some similar implementations let you designate a specific folder for client-side encryption while the rest of your account uses standard server-side encryption. This lets you have AI features for most of your storage and genuine zero-knowledge for specific sensitive files.
On-device AI processing. Some services are moving toward running AI models locally on your device rather than on their servers, which would allow AI-powered features alongside stronger encryption guarantees. This is technically viable on modern phones and computers, though the models available for on-device processing are currently less capable than server-side equivalents.
Local-first apps with optional sync. Applications like Obsidian keep your data local by default and sync across devices through an optional encrypted channel. The server never stores unencrypted content, but you also manage sync and backup yourself.
How daftei Approaches This
daftei uses server-side encryption: files are encrypted at rest with AES-256 and in transit with TLS 1.3. This means daftei can offer AI-powered features — search, organisation, memory assistance — that require server-side access to file contents. It also means daftei is subject to the same trade-off as other server-side encrypted services: the provider can access files if legally compelled to do so.
What server-side encryption at daftei does guarantee: protection against external breaches at the storage layer, strong transit security, and an operational model that doesn’t depend on selling data or training third-party AI models on what you store. daftei doesn’t run advertising, doesn’t sell personal data, and doesn’t use your content to train AI systems that benefit anyone other than you.
The trade-off is clear: daftei chooses AI usability over the strongest privacy guarantee that technical architecture can provide. For users whose primary concern is government access or provider-level surveillance — journalists, lawyers, activists, people in high-risk personal situations — genuine zero-knowledge services are a meaningful consideration. For users who want a private storage experience without AI-assisted features being the enemy of privacy, server-side encryption from a provider with clear data practices is the more practical answer.
The Right Question
“Is this service encrypted?” is the wrong question. Every reputable cloud storage service in use today is encrypted. The question that actually matters is: who holds the keys?
If the provider holds the keys, you trust the provider — their policies, their integrity, their legal jurisdiction, and their future decisions. If only you hold the keys, you trust the mathematics of encryption, which is a very different kind of trust.
Neither answer is universally correct. But you should know which one you’re choosing.