Did you know that the paper contains (assumptions) that are highly idealized to make this attack successful? In fact, the success or failure of the attack depends on these assumptions.

First, what are the assumptions:

They are scenarios of (helpful) hypotheses set by the researcher to reach the desired outcome.
These assumptions may have been tested on a small scale or not tested at all, but they provide a (hypothetical) scenario of what could happen if these assumptions succeed in their ideal form.

The paper itself refers to these assumptions as “benign hardware assumptions.”

Here are the main assumptions in the Google Quantum AI paper:

Physical Error Rate (:

10^{-3} )Approximately one error per 1,000 operations(.

The problem: This rate has only been demonstrated on a small scale )tens or hundreds of qubits like the Willow chip(. When scaled to hundreds of thousands of qubits, correlated errors )correlated errors( and an “error floor” )error floor( appear, which do not decrease as expected, making extrapolation overly optimistic.

Qubit architecture: Superconducting qubits )Superconducting qubits(.

The problem: This architecture is very sensitive to noise, vibrations, and cosmic radiation. When reaching 500,000 qubits, issues with cooling, power, and crosstalk )crosstalk( become very difficult and are not yet practically proven.

Qubit connectivity )Connectivity(: Planar architecture with four-degree connectivity )planar surface architecture, each qubit connected to only 4 others(.

The problem: This limited connectivity increases circuit overhead and slows down execution. Better architectures )like long-range or higher-degree connectivity( are not practically available in current devices and require unproven technologies.

Error correction code: Variants of the surface code )expanded versions of the surface code(.

The problem: It has been successfully tested over small distances )distance 5–7(. For the large distances needed for 500,000 qubits, decoding )error analysis and correction( becomes very slow, and an error floor appears that prevents achieving the required accuracy for the attack in minutes.

Number of physical qubits required: Less than 500,000 physical qubits.

The problem: This number is extrapolated from the Google )Willow( chip.
)extrapolation( from 105 qubits to half a million.

Scaling to this size has not been tested, and engineering challenges )such as maintaining uniform quality across all qubits( make it difficult to achieve in the coming years.

Execution time: The attack can be carried out in minutes )about 9 to 23 minutes(.

The problem: It relies on very fast cycle times and immediate error correction. In reality, as the system grows, decoding and error correction times lengthen, potentially turning minutes into hours or days.

In the end, there are dozens of research papers in the same language as Google that set scenarios based on certain assumptions and produce overly optimistic, unrealistic scenarios about how the attack could be feasible on encryption.

There are papers from 2014 discussing this attack with the same assumptions.

But so far, the “expected quantum computer” that applies this theory to reality has not yet appeared.