Harvest Now, Decrypt Later: Your Long-Shelf Data Is Already at Risk

When I talk about quantum risk with leaders, the most common reaction is to postpone: "the quantum computer doesn't yet exist on this scale, that's a problem for the next decade." The sentence is technically correct and strategically wrong.

It is wrong because there is an attack that does not depend on the machine being ready. He only depends on patience and storage space, two resources that any serious opponent has in abundance. The name is "harvest now, decrypt later": harvest now, decrypt later.

Understanding this mechanism is what separates those who deal with the topic with calibrated urgency from those who leave it until later and discover, late, that "later" started years ago.

The mechanism, without mystique

The idea is simple and uncomfortable. The attacker intercepts and stores encrypted traffic and data today, even without being able to read it now. He bets on the future: when there is a quantum computer capable of breaking the public key encryption used to protect that data, it will decipher what it has stored.

There is no real-time breakdown, there is no alarm going off. Capture is passive. Anyone who observes a network point, operates intermediate infrastructure or compromises data transit can simply copy the encrypted stream and store it.

The attacker's calculation is cold. Storage is cheap and getting cheaper. If a set of data is worth the effort, storing ten years of encrypted traffic is a small investment compared to what you gain from decrypting it. He doesn't need to get the exact date when quantum computing arrives. It just needs to arrive before the data loses value.

Why long validity is the central point

The right question is not “when will the quantum computer exist”. It's "how long this data needs to remain secret." Cross both horizons and you have your exposure.

Think of a window. On the one hand, the lifetime of the confidentiality of the data: how many years will that information still cause damage if it is leaked. On the other, the time until relevant quantum capacity exists. If the validity of the secret exceeds the arrival of the machine, the data captured today will be readable within the period in which it still matters. This data is already, in practice, compromised.

This is why the risk is not uniform. An ephemeral chat session, which loses value in hours, hardly justifies quantum concern. Data that needs to remain protected for a decade or more is already in the critical zone.

Who has data like this? Health, with medical records and genetic history. Financial sector, with long-term contracts and positions. Industry, with intellectual property and process secrets. Defense and diplomacy, with classified information. And the public sector, which stores data on citizens whose sensitivity does not expire: identity, biometrics, tax, judicial and social security records.

Where the exhibition lives today

The vulnerability is concentrated in public-key cryptography. RSA and ECC protect the key exchange that establishes secure connections. It is exactly this trade-off that a capable quantum computer can undo.

The dangerous detail: a lot of traffic is protected by a session key negotiated by this vulnerable mechanism. If the attacker saved the handshake and the encrypted stream, breaking the key exchange in the future delivers the session key, and the session key delivers the content. The symmetric encryption that protected data in transit, such as AES, is not the weak link. The weak link is what established the key.

Hence the focus on connections and data that crossed the network in a way that could be captured. Backups that were transferred, synchronizations between data centers, integrations between systems, everything that was transmitted under vulnerable protection is a candidate for having been harvested.

It is worth separating two scenarios. Data in transit, captureable by anyone observing the network, is the classic case of harvest now. Data at rest protected by strong symmetric encryption has a different risk profile, although poorly protected keys and envelopes can reintroduce the problem. The cryptographic inventory, which I deal with in another article in the series, is what allows you to distinguish one from the other in your reality.

What does this change in your risk assessment

The practical consequence is that the risk clock has already started. Every day that long-term data travels under vulnerable encryption is another day of material potentially archived by someone.

This reorders priorities. Instead of treating post-quantum migration as an item on a future agenda, you start classifying data by the validity of confidentiality. What needs to remain secret for a long time rises to the top, because it is what suffers most from early capture.

It also changes the conversation with suppliers and partners. It's not enough to ask if they encrypt the data. The question becomes which algorithm, and whether there is a plan to transition to post-quantum cryptography. A partner that moves your long-term data under vulnerable protection is a vector of its own risk.

And the compliance stance changes. Under LGPD, protecting personal data by appropriate technical measures is a continuing obligation. When there is a known method of compromising data that you are forced to keep for years, ignoring the issue becomes difficult to sustain in the face of an incident or a question from the regulator.

What to do before migrating everything

You don't need to solve the entire problem at once to significantly reduce risk. Urgency is selective.

The first step is to map data by confidentiality validity. List what needs to remain secret for five, ten, twenty years. This cut reveals where harvest now really hurts and where post-quantum migration has an immediate return.

The second is to understand the path of this sensitive data: where it travels, under what protection, passing through which suppliers. This is where the cryptographic inventory stops being bureaucracy and becomes concrete risk management.

The third is to prioritize the protection of key exchange in these critical flows, adopting, when possible, a hybrid mode that combines the classical algorithm with the post-quantum one. You don't need to migrate the entire park to protect what matters most first.