What is harvest now, decrypt later?

Harvest now, decrypt later means recording encrypted data today to decrypt it once quantum computers are strong enough. Anything encrypted on a public blockchain is stored for good, so it is open to this by design.

Also called store now, decrypt later, steal now, decrypt later, retrospective decryption or HNDL.

Explain it like I’m five

Imagine sending sealed letters through a town where every mail truck has glass walls. Nobody can open the envelopes today, because the lock on each one is too hard to pick. But anybody can photograph every sealed letter and keep the photos in a warehouse. If a master-key machine is invented in ten years, the photographer opens every old envelope at leisure.

In real life the envelope is a connection protected by a key exchange such as RSA or elliptic-curve Diffie-Hellman, the photograph is a copy of the network traffic, and the master-key machine is a quantum computer running Shor’s algorithm. A letter that is still secret by then is lost, and nothing the sender does in the meantime can pull a photographed letter back.

Why it matters

A defense against this threat has to be in place before the data is sent. Traffic that was recorded under a quantum-vulnerable key exchange cannot be protected afterwards, and encrypting it again changes nothing about the copy the attacker already holds. NIST names the threat as one of the main reasons the move to post-quantum cryptography is urgent.

The risk depends on how long a secret must stay secret. A one-time code that expires in a minute is worthless to a future attacker. Health records, identity data, state secrets and the private history of a payment system keep their value for decades.

How it works, step by step

  • Two parties run a key exchange over a public network, such as RSA or elliptic-curve Diffie-Hellman (ECDH), to agree on a secret key.
  • They encrypt the real data with a fast symmetric cipher such as AES under that key (session encryption).
  • An adversary with a cable tap, a compromised router or a server log records the whole exchange, the public values of the handshake and the ciphertext (the harvest). It costs only storage.
  • The recording waits. Nothing has been broken, and the traffic looks safe.
  • Years later a large quantum computer runs Shor’s algorithm on the recorded public values and recovers the secret behind the key exchange (the quantum step).
  • With the session key rebuilt, the attacker decrypts the stored ciphertext with ordinary AES, which was never broken (the decryption).
  • The defense belongs in the first exchange. A post-quantum or hybrid key exchange such as ML-KEM leaves the recording with nothing a quantum computer can open (post-quantum key exchange).

A worked example

NIST states the timing rule from Michele Mosca as X + Y > Z. Suppose data must stay secret for X = 10 years, the migration takes Y = 5 years and a capable quantum computer arrives in Z = 12 years. Then 10 + 5 = 15 is more than 12, so the data is at risk. The migration had to finish within 12 minus 10, which is 2 years, and it takes 5, so it is three years late.

Blockchains show the same pattern. A shielded Zcash note is encrypted on chain with a key that comes from an elliptic-curve exchange, and the ephemeral public key is stored next to the ciphertext. Google Quantum AI’s 2026 whitepaper notes that a quantum attacker who knows the target address can recover that key and decrypt the note, including the amount and the memo, so years of old shielded transactions can lose their privacy in retrospect.

What the threat requires

  • It needs only storage today and a cryptanalytically relevant quantum computer later, and nothing about the live connection looks wrong.
  • It targets confidentiality that rests on public-key key exchange or public-key encryption, such as RSA, Diffie-Hellman and ECDH.
  • NIST judges that approved symmetric ciphers and hashes with at least 128 bits of classical strength still meet security category 1, so the fix concerns the key exchange and leaves AES in place.
  • It cannot be repaired afterwards. Only data protected by a post-quantum or hybrid exchange from the start is safe.
  • Signatures face no such threat, since forging one needs the quantum computer at the time of the attack. NIST says moving them is urgent anyway, because the transition takes long and many parties must coordinate.

Common misunderstandings

  • Harvest now, decrypt later needs no quantum computer today. The harvesting is ordinary storage, and only the decrypting waits for the machine.
  • Bigger AES keys do not help, because the weak link is the key exchange that delivers the key.
  • Encrypting again later does not help, because the attacker already holds the old ciphertext.
  • Bitcoin and Ethereum transactions are not encrypted, so there is nothing to decrypt. Their quantum risk comes from exposed public keys and forged signatures.

Where it matters in practice

  • Cloudflare reported on 28 October 2025 that more than half of the human traffic to its network used post-quantum key agreement and was protected against store-now, decrypt-later.
  • Signal’s PQXDH of 19 September 2023 adds CRYSTALS-Kyber to X25519 and names the threat Harvest Now, Decrypt Later.
  • Apple’s iMessage PQ3 of 21 February 2024 combines Kyber with P-256 elliptic-curve cryptography.
  • OpenSSH 10.0 of 9 April 2025 makes the hybrid mlkem768x25519-sha256 the default key agreement.
  • Google Quantum AI’s whitepaper warns that a quantum attacker could deanonymize years of past transactions for known addresses on privacy chains such as Zcash, Monero and Litecoin’s Mimblewimble.

History

In a 2015 note on whether cybersecurity is ready for quantum computers, Michele Mosca argued that systems needing long-term confidentiality must migrate first. NIST now states his rule as Mosca’s theorem. Signal named the threat in 2023, and NIST’s IR 8547 of October 2026 calls it one of the main reasons the transition is urgent.

In Redacted

Nothing Redacted writes on chain is encrypted, so there is nothing stored there for a future computer to decrypt. In bunker mode your account is 25 words you keep, never derived from a wallet key, so a broken wallet key does not open your private balance.

Bunker mode →

Related terms

Sources: Transition to Post-Quantum Cryptography Standards, NIST IR 8547, October 2026 · Quantum resistance and the Signal protocol, PQXDH, Signal, 19 September 2023 · State of the post-quantum Internet in 2025, Cloudflare, 28 October 2025 · Securing elliptic curve cryptocurrencies against quantum vulnerabilities, Babbush et al., March 2026 · Cybersecurity in an era with quantum computers: will we be ready?, Mosca, 2015

Updated November 1, 2026. All terms