AI Becomes a New Cryptographic Threat: Mathematicians Warn of 'Severe Upheaval'

Cryptographic security, which has relied on mathematical algorithms for decades, is facing a threat for the first time that is unrelated to quantum computers. According to industry specialists, the successes of artificial intelligence running on classical computing architecture in solving complex mathematical problems make this progress a comparably significant threat to data protection. This is reported by the Financial Times.

From Quantum to Classical: Shifting Priorities

Until recently, the main potential problem for cryptographers was considered to be the emergence of sufficiently powerful quantum computers using subatomic particle physics to solve problems inaccessible to classical machines. The timelines for creating such devices were traditionally estimated in years. However, AI models running on familiar computing infrastructure have advanced so far in number theory and algorithm development that they have begun solving problems previously considered too complex. This turns the threat into a more tangible and near-term scenario than the quantum one.

Conflict Surrounding OpenAI

According to the Financial Times, concern in the scientific community has been heightened by the ability of OpenAI models to solve mathematical problems that had eluded scientists for a long time. These successes caused discontent among a specialized association, which urged mathematicians to sever ties with the company. In response, OpenAI donated five million dollars to a specialized fund to smooth over the conflict.

'Very Serious Upheaval'

Australian mathematician Jordy Williamson notes that these AI systems are undergoing a 'complete restructuring of the notion of complexity' upon which modern cryptography is built. 'I think we are in for upheaval in the future, very serious upheaval,' he stated. Currently, AI has not managed to crack the primary method of cryptography, but mathematicians warn that the probability of such a breakthrough is growing. In particular, the widely used method based on multiplying two very large prime numbers has remained effective because the reverse process—determining the original numbers—is difficult for computers.

Reaction of Regulators and Business

The UK Financial Conduct Authority told the Financial Times that it is testing a new supervisory tool across 50 major organizations to assess preparedness for technological threats in cryptography related to both AI and quantum computing. Next month, the regulator intends to share practical recommendations with more than 35,000 companies providing financial services within the country's jurisdiction. Professor of cryptomathematics at the University of Manchester Daniel Shiu believes that for 'more mathematical' forms of cryptography, the threat of AI cracking is already quite tangible, and vulnerabilities are being discovered in less mature algorithms designed to protect against quantum attacks.

Revision of Standards

Authorities in several countries, including the US, China, and Germany, have begun revising existing and proposed cryptographic methods to account for new types of threats. Telecommunications giant BT announced the start of preparations for migration to new cryptographic methods that take into account future threats from quantum computers, focusing on ensuring faster adaptation of cryptographic methods as standards, technologies, and the threats themselves evolve.