Cybersecurity researchers received unfavorable news last week. According to research published in recent articles by Google and Oratomic, a quantum computing firm, quantum computers that may crack the encryption algorithms protecting the internet might appear sooner rather than later.
Bas Westerbaan, a cybersecurity researcher at Cloudflare, which protects a sizable portion of the internet, told, “It’s a real shock.” “Our efforts will need to be accelerated significantly.” Cloudflare declared on Tuesday that it was “accelerating” its timetable to get ready for quantum computers to 2029.
The authors of the research told that AI was “instrumental” in creating the algorithm used by the Oratomic team. One of the authors of the research, Dolev Bluvstein, states, “There is no doubt that we used AI to accelerate this development.” “There is absolutely no doubt.”
Quantum bits, or “qubits,” are the building blocks of quantum computers. They employ the paradoxical principles of quantum mechanics to carry out some calculations far more quickly than is feasible with conventional computers. Internet security is at risk due to its speed. Everything, even top-secret documents and WhatsApp conversations, depends on the fact that it would take the most powerful supercomputer far longer than the universe’s lifespan to crack their encryption and reveal their contents to the public. On the other hand, a quantum computer might theoretically complete the same task in a few of days.
Today’s quantum computers are too small to pose a threat, but according to a 2025 survey, there is a 39% probability that this could change within the next ten years as they become more potent and their algorithms become more effective, necessitating ever-tinier quantum computers to crack encryption. The National Institute for Standards and Technology (NIST) in the United States has set a deadline of 2035 to get ready for their arrival. According to many quantum computing experts who talked with TIME, Oratomic and Google’s findings could “significantly” reduce the time needed to construct a quantum computer that poses a danger to encryption.
According to Bluvstein, who recently co-founded Oratomic with the goal of creating the first practical quantum computer, “those in the know will be like: ‘oh s—, it’s coming.” “I believe that the world is not ready right now.”
Many of the authors’ assumptions are “untested,” according to Jeff Thompson, an associate professor at Princeton and CEO of the atomic quantum computing startup Logiqal. The research has not yet undergone peer review. Thompson continues, “If you just assume better qubits, it’s very easy” to reduce the computer’s size.
On March 25, the week before the Google and Caltech studies were published, Google established a plan for securing its systems against quantum computers by 2029—six years before the NIST deadline of 2035.
AI leaders have consistently stated that AI will boost scientific advancement. In 2025, Sam Altman predicted that AI’s ability to accelerate scientific advancement will significantly improve quality of life. Beyond breaching encryption methods, quantum computing experts expect that the new technology may aid in physics discoveries as well as the development of novel pharmaceuticals and materials. They may one day contribute to the development of more powerful and efficient AI models. However, Westerbaan believes that developing a quantum computer before transitioning to post-quantum encryption could result in data leaks, blackmail, and the shutdown of businesses.
According to Westerbaan, “nearly every system in the world becomes completely vulnerable to a quantum attacker.”
“This whole thing would not work” without AI
Cosmic rays from the sun and other environmental disturbances can readily break qubits. Redundancy, or distributing information among numerous qubits, has been the solution, allowing the computer to continue operating even in the event that some qubits fail. This makes a quantum computer more reliable, but at the expense of controlling a lot more little particles.
A single qubit in atomic quantum computers—quantum computers with qubits composed of atoms—may require 100–1,000 atoms to encode. However, the Oratomic researchers’ approach reduces the number of particles needed to construct an atomic quantum computer by 100 times, requiring only three atoms to encode a qubit.
According to Robert Huang, one of the authors of the research, the team’s main algorithms performed “about 1,000 times worse” at first. “This wouldn’t work at all.”
Huang made the decision to try optimizing the algorithms using OpenEvolve, an open-source program that uses LLMs like Google’s Gemini and Anthropic’s Claude, in a manner similar to natural selection. He says, “I didn’t think you would find anything useful.”
He was astonished. The AI incorporated previous scientific achievements in a “novel way,” exhibiting awareness of narrow sub-disciplines in quantum computing while testing thousands of various hypotheses. Without the AI, he believes he and his team would have tried a few concepts, discovered they didn’t work, and concluded that “the whole thing is not possible.” The AI approaches dramatically improved the performance of several of the paper’s most essential algorithms.
The paper’s author, John Preskill, who is regarded as a pioneer of quantum computing, adds, “I’m surprised by how much we were able to reduce the qubit count.” He pointed out that people continued to be the main forces behind the study, “asking the right questions and then guiding the AI towards answers that are useful and informative.”
Before feeling comfortable revealing the algorithm that the AI had developed, Huang and his colleagues spent months validating it. The scientists stress that “many open challenges” still need to be overcome before a dangerous quantum computer is constructed.
“This has social ramifications.”
Bluvstein claims that before the paper was published, members of the Oratomic team briefed U.S. government authorities. He laughs and continues, “This is the first paper I have, at least personally, ever written where I’m like, ‘Wow, this has implications for society.” “For us, this is a new regime.”
According to Scott Aaronson, an independent quantum computing researcher at UT Austin, the National Security Agency (NSA) and NIST are the U.S. organizations that would normally have input on the publication of such a work. Regarding the agencies his team had contacted, Bluvstein declined to comment.
Given the possible ramifications of developing a quantum computer, the Oratomic team carefully considered whether aspects of their research should be published. Although Bluvstein states that the team intends to publish a follow-up work outlining its use of AI, the paper draft does not address the use of AI in obtaining its important discoveries.
There have been other interested parties than U.S. government officials. Huang, who co-founded Oratomic with some of the paper’s co-authors after leaving Google Quantum AI in 2026 to work at Caltech, informed a buddy at Google’s quantum program that he had been using AI and “seeing lots of crazy results.” Google hired a quantum researcher to create AI-based “discovery pipelines” in early March, a few months later.
Less than a week before to the simultaneous Google and Oratomic publications, on Tuesday, March 24, Google revealed a new internal atomic quantum computing project. Atoms, superconducting circuits, and photons can all be used in the development of quantum computers. According to Umesh Varizani, a quantum computing researcher at UC Berkeley, the “million-dollar question” is what kind of quantum computer will be the simplest to construct.
The Oratomic paper makes atomic quantum computers more efficient, but it has no effect on the resources needed to create other kinds of quantum computers. Since 2014, Google has been working on superconducting quantum computers, which are seen to be the more promising approach. In 2024, Google invested in QuEra, a company that produces atomic quantum computers.
In an email comment, a Google representative claimed that the business has been evaluating atomic quantum computing for “many years” and has published research on the use of AI for quantum error correction “for years as well,” noting that some “newer entrants are now pursuing similar ideas.”
According to Bluvstein, “our research over the past half year has been surprising.” “Our research and partnerships are having an impact on the industry as a whole.”
