AI Helped Spark a Quantum Breakthrough. The World 'Is Not Prepared'
· Time
Last week, cybersecurity researchers woke up to bad news. Research in new papers published by Google and a quantum computing startup, Oratomic, suggests that quantum computers capable of breaking the encryption protocols that secure the internet may arrive sooner than expected.
“It's a real shock,” Bas Westerbaan, a cybersecurity researcher at Cloudflare, which secures a significant fraction of the internet, told TIME. “We’ll need to speed up our efforts considerably.” On Tuesday, Cloudflare announced that it was “accelerating” its deadline to prepare for quantum computers to 2029.
Visit turconews.click for more information.
AI was “instrumental” in developing the Oratomic team’s algorithm, the paper’s authors tell TIME. “There is no question that we used AI to accelerate this development,” says Dolev Bluvstein, one of the paper’s authors. “No question at all.”
Quantum computers are built from quantum bits, or “qubits,” which use the counterintuitive laws of quantum mechanics to perform certain calculations much faster than is possible with ordinary computers. That speed poses a threat to internet security. Everything from WhatsApp messages to top-secret documents rely on the fact that it would take the most powerful supercomputer much longer than the age of the universe to break their encryption and expose their contents to the world. A quantum computer, however, could theoretically do the same work in days.
Today’s quantum computers are too small to be dangerous, but a 2025 survey found a 39% chance of this changing in the next decade, as quantum computers grow more powerful and the algorithms they run become more efficient—requiring ever smaller quantum computers to break encryption. The U.S. National Institute for Standards and Technology (NIST) has set a 2035 deadline to prepare for their arrival. Together, Oratomic and Google’s results could “significantly” shorten the development time of a quantum computer that threatens encryption, according to multiple quantum computing experts who spoke to TIME.
“People in the know will be like: ‘oh s—, it's coming,’” says Bluvstein, who recently co-founded Oratomic, which aims to build the first useful quantum computer. “The world is currently, in my view, not prepared.”
The paper has not yet been peer-reviewed, and many of the assumptions that the authors make are “untested,” says Jeff Thompson, an associate professor at Princeton and CEO of atomic quantum computing startup Logiqal. It's “very easy” to reduce the size of the computer “if you just assume better qubits,” Thompson adds.
On March 25, the week before the publication of the Google and Caltech papers, Google announced a timeline to secure its systems against quantum computers by 2029—six years before NIST’s 2035 deadline.
AI leaders have repeatedly promised that AI would accelerate scientific progress. “The gains to quality of life from AI driving faster scientific progress … will be enormous,” wrote Sam Altman in 2025. Beyond cracking encryption protocols, quantum computing researchers hope the new technology could help make discoveries in physics, and design new drugs and materials. Someday, they may help run more powerful and efficient AI models. But according to Westerbaan, the development of a quantum computer before the transition to post-quantum encryption could lead to data leaks, extortion, and businesses being taken offline.
“Almost every system in the world becomes vulnerable altogether to a quantum attacker,” Westerbaan says.
Without AI, ‘this whole thing would not work’
Qubits are easily broken by perturbations from the environment, such as cosmic rays from the sun. The solution has been redundancy: spreading information across many qubits, so that the computer still works even if some qubits fail. This increases the reliability of a quantum computer, at the cost of having to control many more tiny particles.
In atomic quantum computers—quantum computers with qubits made out of atoms—it can take 100 to 1,000 atoms to encode a single qubit. But the algorithm found by the Oratomic researchers requires just three atoms to encode a qubit, reducing the number of particles required to build an atomic quantum computer by 100 times.
Initially, the performance of the team’s key algorithms was “about 1,000 times worse,” says Robert Huang, one of the paper’s authors. “This whole thing would not work.”
Huang decided to try using OpenEvolve, an open-source tool which harnesses LLMs such as Google’s Gemini and Anthropic’s Claude, to optimize the algorithms in a process analogous to natural selection. “I didn't expect you would find anything useful,” he says.
He was surprised. The AI combined past scientific results in a “novel way,” demonstrating understanding of niche sub-disciplines in quantum computing as it tried thousands of different ideas. Without the AI, he says, it’s likely that he and his team would have tried a few ideas, seen that they didn’t work and decided that “the whole thing is not possible.” The AI’s proposals ended up significantly improving the performance of some of the most important algorithms in the paper.
“I'm surprised by how much we were able to reduce the qubit count,” says John Preskill, an author on the paper who is widely considered to be a pioneer of quantum computing. He noted that humans were still the primary drivers of the research, “asking the right questions and then guiding the AI towards answers that are useful and informative.”
Huang and his colleagues spent months verifying the algorithm that the AI had derived before they felt confident sharing it. Still, the authors emphasize, “many open challenges” remain before a dangerous quantum computer is built.
‘This has implications for society’
Members of the Oratomic team briefed U.S. government officials prior to the paper’s publication, according to Bluvstein. “This is the first paper I have, at least personally, ever written where I'm like, ‘Wow, this has implications for society,’” he says with a laugh. “It's a new regime for us.”
The U.S. agencies that would typically have input on the publication of such a paper are the National Security Agency (NSA) and NIST, according to Scott Aaronson, an unaffiliated quantum computing researcher at UT Austin. Bluvstein declined to comment on which agencies his team had spoken to.
The Oratomic team thought carefully about which parts of their methodology were responsible to publish, given the potential consequences of a quantum computer being developed. The paper draft does not mention the use of AI in deriving its key results, but Bluvstein says that the team plans to publish a follow-up work detailing its use of AI.
U.S. government officials have not been the only interested party. Huang—who previously worked at Google Quantum AI and left in 2026 to work at Caltech and co-found Oratomic with some of the paper co-authors—told a friend at Google’s quantum initiative that he had been using AI and “seeing lots of crazy results.” A few months later, in early March, Google posted a job for a quantum researcher to develop AI-based “discovery pipelines.”
On Tuesday, March 24, less than a week before the concurrent Google and Oratomic publications, Google announced a new internal atomic quantum computing initiative. Quantum computers can be developed in many different ways, including using atoms, superconducting circuits and photons. Which type of quantum computer will be easiest to build is the “million-dollar question,” says Umesh Varizani, a quantum computing researcher at UC Berkeley.
The Oratomic paper increases the efficiency of atomic quantum computers, but does not affect the resources required to build other types of quantum computers. Google has worked on superconducting quantum computers, which have been seen to be the more promising approach, since 2014, and invested in QuEra, which makes atomic quantum computers, in 2024.
A Google spokesperson said in an emailed statement that the company has been assessing atomic quantum computing for “multiple years,” and has published research on using AI for quantum error correction “for years as well,” adding that some “newer entrants are now pursuing similar ideas.”
“Our research over the past half year has been surprising,” says Bluvstein. “We are indeed seeing the effects of our research and collaborations on the broader industry landscape.”