AI Vulnerability Surge Outpaces Calls for Development Slowdown
While AI firms discuss a possible industry-wide pause, researchers report a rapid increase in security flaws uncovered by widely used AI chatbots.
Industry leaders are debating a coordinated slowdown of artificial-intelligence development, but security researchers say the most immediate risk comes from a growing number of vulnerabilities being exposed by AI tools. A recent investigation highlighted nine flaws in ATM encryption and authentication software, illustrating how AI-driven analysis can reveal weaknesses across the software supply chain.
In a separate experiment, an open-source AI model stripped of its safety constraints identified security gaps in household devices and a personal computer, then offered recommendations for strengthening defenses. OpenAI plans to release an AI model with advanced cyber capabilities to a limited set of partners, giving them time to address potential threats.
The discussion of a slowdown has also raised legal questions, with AI executives wondering whether antitrust regulations might impede collective action. Meanwhile, policymakers in the United Kingdom have expressed concern over the rapid emergence of autonomous AI agents and their possible societal impacts.
"We need to understand the legal implications before any coordinated effort can be implemented," said an OpenAI spokesperson.
"The pace at which vulnerabilities are being discovered underscores the urgency of improving security measures," a security researcher involved in the ATM study commented.
Despite differing views on a development pause, experts agree that the immediate challenge lies in securing systems against AI-identified flaws before they can be exploited.