SINGAPORE/HONG KONG, April 30 (Reuters) – Banks across Asia are tightening their scrutiny of artificial intelligence tools as the latest models raise concerns about potential cyber threats. This heightened vigilance is particularly evident in Singapore, one of the region’s largest financial hubs, where institutions worry that hackers could exploit these new technologies to discover vulnerabilities more swiftly and execute broader cyberattacks.
In a notable development, AI startup Anthropic has unveiled Claude Mythos Preview as part of its Project Glasswing, a restricted-access cybersecurity initiative. Major financial institutions such as JPMorgan Chase are identified as launch partners, while others, including Goldman Sachs, Citigroup, Bank of America, and Morgan Stanley, are either testing the model or have access to it, according to sources familiar with the matter.
Mythos stands out as Anthropic’s most advanced AI model, specifically engineered for defensive cybersecurity operations. However, its introduction has raised alarms within the industry, with the company stating that the tool has identified “thousands” of significant vulnerabilities across “every major operating system and web browser.” This revelation has amplified concerns about the potential ramifications for traditional software security.
On Thursday, Australia’s prudential regulator issued a warning that banks are lagging behind in adapting to rapid advancements in AI technology. In response to these developments, DBS Group CEO Tan Su Shan remarked that tools like Mythos heighten existing risks in the financial sector. “What Mythos does is it amplifies the risk,” she stated during a briefing following the bank’s first-quarter earnings, which surpassed expectations. She highlighted that the speed at which vulnerabilities can be exploited has increased, stating, “From both a speed perspective, it’s faster to market, and from a volume perspective, the blast radius is fast.”
Tan acknowledged that while attackers could utilize AI tools to spot weaknesses more quickly, banks can also leverage similar technologies to bolster their defenses. Despite the risks, she emphasized that DBS views AI as a “net positive,” pointing to benefits in coding and operational efficiencies.
Meanwhile, the Oversea-Chinese Banking Corporation (OCBC), Singapore’s second-largest lender, has voiced its commitment to responsible AI implementation, emphasizing that all solutions undergo stringent assessments before deployment. “All such solutions undergo rigorous assessment and validation before deployment,” stated Praveen Raina, OCBC’s head of group operations and technology.
Similarly, United Overseas Bank (UOB), the third-largest lender in Singapore, affirmed that while AI is a fundamental component of its digital transformation strategy, its utilization is strictly governed by existing cybersecurity protocols and internal guidelines. A spokesperson for UOB remarked, “We take a disciplined and responsible approach to innovation.”
In a broader context, Standard Chartered CEO Bill Winters referred to Anthropic’s Mythos as a “sensational representation” of the increasing cyber risks that financial institutions face. However, he assured that Standard Chartered is well-prepared to tackle these challenges.
The growing focus on AI in cybersecurity highlights a pivotal moment for the banking sector, as institutions must navigate the dual-edged nature of this technology. While AI presents opportunities for enhanced security and operational efficiency, it also introduces new threats that could disrupt the financial landscape. As banks implement stricter controls and assessments, the industry is poised for significant changes in how it approaches both AI adoption and cybersecurity in the years to come.
