Recently, DeepMind CEO Demis Hassabis made a statement that deserves serious attention from the cryptography community.
He suggested that artificial intelligence, even when run on classical computing systems, may be capable of finding solutions to most problems that were previously assumed to require quantum computing.
Executive Chairman of Post–Quantum.
As evidence, he pointed to the unexpected success of AlphaFold, which demonstrated that machine learning could master complex molecular modelling once believed to demand quantum-scale simulation.
For cryptographers, this is not just a theoretical curiosity. It directly challenges assumptions about the pace and nature of the threats facing today’s public key encryption.
If AI can achieve quantum-like breakthroughs without requiring quantum hardware, then the threat horizon shifts significantly closer.
The quantum threat
Quantum computing will one day be able to break classical encryption methods, such as RSA, because it can efficiently solve mathematical problems that are difficult for classical computers.
For example, RSA uses a number derived from the product of two large primes as its public key, and uses another number derived from the prime factors as its private key.
Because finding the prime factors from a large number is computationally difficult for classical computers, especially when dealing with very large numbers, the private key remains secure.
If rapid factorization of large numbers were possible, it would mean that the private keys could be discovered from the public keys, rendering virtually all data and communications vulnerable.
Quantum computers process information using quantum bits (qubits) instead of traditional computer bits. While classical computers use bits that are either 0 or 1, qubits can represent both states simultaneously.
This capability makes quantum computers exceptionally powerful for solving complex problems such as factorization, essentially calculating in parallel to solve such problems.
Why AI may threaten encryption before quantum
AI run on classical systems will not have the same level of computing power as a quantum computer. However, just because it cannot solve these problems directly, this does not mean that it will not be instrumental in breaking traditional cryptography.
To borrow a chess analogy: Many of the best players in the world now use AI to refine and improve their game. In fact, there is significant competition among top players for access to the newest models like Leela Zero and those developed by Google DeepMind.
These AI models act as training partners, coaches, and research assistants, helping the world’s top players to think about positions in a different way.
For example, top players will frequently use the engines to identify novelties they wouldn’t spot on their own – perhaps sacrificing a rook for a less powerful minor piece in the opening to achieve a better overall position.
The AI augments the talents of the human players, helping them think in different ways to identify new solutions and approaches they can deploy in high-level games.
In the case of cryptography, AI may one day play a similar supporting role. Used alongside human experts, AI can comprehensively search research papers, journals and databases to find and combine relevant work.
This may have the potential to accelerate the process of finding solutions by narrowing down the search space and providing unorthodox ideas.
In a world where most people are experts in a single field, the operational research based, cross-disciplinary approach of AI may accelerate solutions to the complex problems on which traditional cryptography is based.
Although all eyes are on Shor’s algorithm being the one and only solution to crack RSA, there are already labs which do anything but that. They use AI to scour decades or even centuries of academic research, hypotheses and conjectives which are used in different fields.
Just like in chess, a talented code breaker armed with the lateral thinking of a powerful AI engine might just stumble across a combination of pre-existing knowledge that helps them crack RSA – without the need for a quantum machine.
The limits and the promise of AI
Of course, there are limitations. AI has not yet demonstrated an ability to crack cryptographic keys and there remain questions around the ability of AI to produce precise answers due to its reliance on numerous parameters and potential for hallucination.
These drawbacks mean that AI cannot currently be trusted to independently complete high-accuracy mission-critical tasks. But AI’s rapid acceleration means complacency isn’t an option. The technology is already freely available for use and misuse.
Unlike quantum computing, which faces steep hardware challenges, AI research can be iterated, trained, and scaled on existing infrastructure. In practical terms, this means AI might pose a practical threat to encryption before quantum computers.
We must also reckon with the fact that these are not mutually exclusive technologies. AI and quantum computing may evolve in parallel or even intersect.
AI could help optimize quantum computers, or quantum-enhanced AI could become a new paradigm altogether. That overlap could bring forward the timeline for when encryption becomes genuinely vulnerable.
The time to prepare is now
In this context, quantum-safe cryptography becomes more than just a long-term strategy – it is an urgent necessity.
Quantum-safe cryptography is not only a defense against future quantum computers, but these new, more robust, encryption algorithms are also likely to thwart AI-powered attacks as well.
For too long, the conversation around cybersecurity has treated the cryptographic migration as a problem to be solved in a future decade.
Governments and businesses cannot afford to wait for a definitive timeline – especially when estimates from the NCSC and other government bodies suggest that migration to new encryption algorithms is likely to take over a decade.
The tools to mitigate these risks are already here, and the cost of inaction will only grow. The age of quantum-safe cryptography isn’t coming – it’s already here.
We’ve featured the best ransomware protection.
This article was produced as part of TechRadarPro’s Expert Insights channel where we feature the best and brightest minds in the technology industry today. The views expressed here are those of the author and are not necessarily those of TechRadarPro or Future plc. If you are interested in contributing find out more here: https://www.techradar.com/news/submit-your-story-to-techradar-pro
