Government security experts are warning that cybercriminals could possess the ability to exploit bank accounts using artificial intelligence within the next three to five months, according to reports. Officials working with Britain’s AI Security Institute (AISI) have expressed concerns that increasingly advanced AI models, particularly those developed in China with fewer safety restrictions, could enable widespread fraud.
These AI models reportedly allow developers to easily remove safety features that would typically prevent the misuse of AI for criminal activities, such as accessing personal financial information or company data. This contrasts with Western AI models that often have guardrails to block illicit requests. The fear is that such accessible technology could soon be in the hands of hackers globally.
Banks may struggle to adequately defend against these sophisticated attacks, with some technology advancing so rapidly that financial institutions might only have hours to prevent breaches, according to a paper published by the Financial Stability Institute. Ministers are reportedly concerned about the pace of AI development and anticipate a surge in complex cyberattacks.
The United States has also seen concerns rise regarding AI's potential dangers. Leading AI firm Anthropic recently barred Britain's AI Safety Institute from testing its latest model. Anthropic reported thwarting five attempts this year to use its model for bioweapon creation, highlighting the dual-use nature of advanced AI research. The company's 'Claude' model, while capable of scientific breakthroughs, could also be used to engineer dangerous pathogens.
These developments come amid a global debate on AI safety, intensified by a researcher's departure from a major AI company over fears of an uncontrolled AI arms race. The former researcher issued a stark warning about the potential for AI to cause existential harm, increasing calls for more rigorous AI safety measures. AI safety advocates are urging governments to publicly acknowledge and address the imminent risks associated with the technology.
A government spokesperson stated that while AI is enhancing some cyber capabilities, businesses are not without recourse. They emphasized the need for all firms to urgently improve their cyber defenses and pointed to existing guidance, such as the Cyber Essentials scheme and recommendations from the National Cyber Security Centre, for managing security at a board level.