Recent developments in artificial intelligence (AI) have raised concerns among cybersecurity experts and regulators about the security of cryptographic tools that protect digital data. Advances in AI-powered mathematical problem solving have prompted calls for organizations to reassess and strengthen the encryption methods used to safeguard sensitive information online.
The UK’s Financial Conduct Authority (FCA) has intensified warnings to financial institutions and other companies about potential vulnerabilities in current cryptographic systems. The regulator has begun testing a new supervisory framework with 50 major institutions to evaluate their readiness against emerging technological risks linked to AI and quantum computing. It plans to issue formal guidance next month aimed at helping the more than 35,000 financial services firms it oversees to prepare for a “quantum-secure future.”
Concerns escalated after OpenAI released hundreds of mathematical papers, some containing purported solutions to longstanding problems, generated with the assistance of its AI models. This move sparked criticism from the Association for Human Mathematics, which accused OpenAI of threatening the integrity of the field and called for mathematicians to cease collaboration with the AI developer. OpenAI has defended its publication strategy and pledged $5 million to support a mathematics-focused foundation.
Experts caution that AI’s increasing proficiency in algorithm design and number theory—the study of properties of whole numbers—could soon challenge the foundations of modern cryptography. Australian mathematician Geordie Williamson described the rapid development as a “complete kind of restructuring of what difficulty is,” warning of potentially “very unsettling major shocks” to current security assumptions.
While AI has yet to break core cryptographic methods, and some cybersecurity professionals remain skeptical of such a breakthrough in the immediate future, the improved problem-solving capabilities of AI have already influenced cybersecurity efforts. For example, AI models developed by OpenAI and Google recently contributed to identifying critical vulnerabilities in PHP, a widely used website programming language, prompting a swift global response.
Traditionally, the looming threat to cryptography has been attributed primarily to the eventual advent of quantum computing—technology that exploits subatomic particle physics to perform calculations far beyond classical computers’ abilities. However, most experts agree that fully operational quantum computers capable of breaking encryption remain years away. Nonetheless, the rapid pace of AI advancements now complicates the risk landscape, prompting regulators and institutions to emphasize proactive measures.
“The evolving nature of cyber threats means firms cannot afford to postpone action,” the FCA said, urging companies to thoroughly assess their current protections and begin implementing strategies to withstand future AI and quantum-related vulnerabilities.