Banks in Asia Tighten Defenses as Frontier AI Exposes Thousands of System Vulnerabilities
2 Sources
[1]
Banks in Asia brace for complex cyber threats from frontier AI
SINGAPORE/HONG KONG, April 30 (Reuters) - Banks across Asia, including those in Singapore - among the region's largest - are tightening checks on artificial intelligence tools as the very newest models have raised fears that hackers could find weak spots faster and launch wider cyberattacks. Anthropic has launched Claude Mythos Preview under Project Glasswing, a restricted-access cybersecurity programme. JPMorgan Chase (JPM.N), opens new tab is a publicly named launch partner, while Goldman Sachs (GS.N), opens new tab, Citigroup (C.N), opens new tab, Bank of America (BAC.N), opens new tab and Morgan Stanley (MS.N), opens new tab have access to, or are testing, the model, Reuters reported, citing sources and company executives. Mythos is Anthropic's most advanced AI model to date, designed for defensive cybersecurity tasks. Its capabilities have sparked fears about the threat to traditional software security after the AI startup said the preview had uncovered "thousands" of major vulnerabilities in "every major operating system and web browser." On Thursday, Australia's prudential regulator warned that banks were not keeping pace with AI developments. "What Mythos does ... is it amplifies the risk," Singapore's biggest bank DBS Group (DBSM.SI), opens new tab CEO Tan Su Shan told reporters in a briefing after the lender posted first-quarter earnings that beat expectations. "From both a speed perspective, it's faster to market, and from a volume perspective, the blast radius is fast," she added. Tan said attackers could use such tools to find weaknesses faster but banks could also use them to defend faster. DBS still sees AI as "a net positive", she said, citing gains in coding and operations. Oversea-Chinese Banking Corporation (OCBC.SI), opens new tab, Singapore's second-largest lender, said it backed responsible AI use after strict checks. "All such solutions undergo rigorous assessment and validation before deployment," Praveen Raina, OCBC's head of group operations and technology, said in a statement to Reuters on Thursday. United Overseas Bank (UOBH.SI), opens new tab, the city-state's third-largest lender, said that while AI was a "strategic pillar" in its digital transformation journey, its use is governed by "existing cybersecurity controls and clear internal guardrails". "We take a disciplined and responsible approach to innovation," a UOB spokesperson said in a statement to Reuters on Thursday. Meanwhile, Standard Chartered Chief Executive Bill Winters told reporters on Thursday that Anthropic's Mythos is a "sensational representation" of a broader trend of rising cyber risks, though he said the bank is well prepared. "Is this a particularly unusual time? No, I don't think so. I think we've had this this threat has been present for for many years," he said. "Just the level of sophistication has increased." Reporting by Rae Wee and Yantoultra Ngui in Singapore and Selena Li in Hong Kong; Editing by Hugh Lawson Our Standards: The Thomson Reuters Trust Principles., opens new tab * Suggested Topics: * Cybersecurity * Data Privacy * Regulatory Oversight Yantoultra Ngui Thomson Reuters Yantoultra Ngui is the Southeast Asia Deals Correspondent of Reuters in Singapore, covering M&A and capital market activities in a region that is fast emerging as one of the world's biggest economies. He previously was a reporter at Bloomberg and The Wall Street Journal (WSJ). Notably, he was part of WSJ's team that covered the financial scandal at Malaysian state fund 1MDB, and that won SOPA Excellence in Breaking News award for the coverage of the assassination of Kim Jong Nam, the half-brother of North Korea's leader Kim Jong Un, in Malaysia in 2018. Yantoultra graduated with an MBA in Finance from Universiti Putra Malaysia (UPM) in 2010.
[2]
Banks in Asia brace for complex cyber threats from frontier AI
SINGAPORE/HONG KONG, April 30 (Reuters) - Banks across Asia, including those in Singapore - among the region's largest - are tightening checks on artificial intelligence tools as the very newest models have raised fears that hackers could find weak spots faster and launch wider cyberattacks. Anthropic has launched Claude Mythos Preview under Project Glasswing, a restricted-access cybersecurity programme. JPMorgan Chase is a publicly named launch partner, while Goldman Sachs, Citigroup, Bank of America and Morgan Stanley have access to, or are testing, the model, Reuters reported, citing sources and company executives. Mythos is Anthropic's most advanced AI model to date, designed for defensive cybersecurity tasks. Its capabilities have sparked fears about the threat to traditional software security after the AI startup said ?the preview had uncovered "thousands" of major vulnerabilities in "every major operating system and web browser." On Thursday, Australia's prudential regulator warned that banks were not keeping pace with AI developments. "What Mythos does ... is it amplifies the risk," Singapore's biggest bank DBS Group CEO Tan Su Shan told reporters in a briefing after the lender posted first-quarter earnings that beat expectations. "From both a speed perspective, it's faster to market, and from a volume perspective, the blast radius is fast," she added. Tan said attackers could use such tools to find weaknesses faster but banks could also use them to defend faster. DBS still sees AI as "a net positive", she said, citing gains in coding and operations. Oversea-Chinese Banking Corporation, Singapore's second-largest lender, said it backed responsible AI use after strict checks. "All such solutions undergo rigorous assessment and validation before deployment," Praveen Raina, OCBC's head of group operations and technology, said in a statement to Reuters on Thursday. United Overseas Bank, the city-state's third-largest lender, said that while AI was a "strategic pillar" in its digital transformation journey, its use is governed by "existing cybersecurity controls and clear internal guardrails". "We take a disciplined and responsible approach to innovation," a UOB spokesperson said in a statement to Reuters on Thursday. Meanwhile, Standard Chartered Chief Executive Bill Winters told reporters on Thursday that Anthropic's Mythos is a "sensational representation" of a broader trend of rising cyber risks, though he said the bank is well prepared. "Is this a particularly unusual time? No, I don't think so. I think we've had this this threat has been present for for many years," he said. "Just the level of sophistication has increased." (Reporting by Rae Wee and Yantoultra Ngui in Singapore and Selena Li in Hong Kong; Editing by Hugh Lawson) By Yantoultra Ngui, Rae Wee and Selena Li
Share
Copy Link
Major banks across Asia are intensifying checks on artificial intelligence tools after Anthropic's most advanced AI model uncovered thousands of vulnerabilities in major operating systems and web browsers. The Claude Mythos Preview has sparked concerns that hackers could exploit frontier AI to identify system vulnerabilities faster and launch larger-scale cyberattacks, prompting Singapore's largest lenders and regional banks to strengthen their cybersecurity controls.
Frontier AI Raises Complex Cyber Threats for Banks in Asia
Banks in Asia are responding to mounting cyber threats from frontier AI by tightening their oversight of artificial intelligence tools, as the latest advanced AI models demonstrate capabilities that could fundamentally alter the cybersecurity landscape. The catalyst for this heightened vigilance is Anthropic's Claude Mythos Preview, launched under Project Glasswing, a restricted-access cybersecurity programme designed for defensive cybersecurity tasks
1
. The AI startup revealed that the preview had uncovered thousands of major vulnerabilities in every major operating system and web browser, raising alarms about how hackers might exploit similar tools to identify system vulnerabilities and launch larger-scale cyberattacks2
.JPMorgan Chase serves as a publicly named launch partner for the programme, while Goldman Sachs, Citigroup, Bank of America and Morgan Stanley have access to or are testing the model, according to Reuters sources and company executives. Meanwhile, regulators are sounding warnings. On Thursday, Australia's prudential authority cautioned that banks were not keeping pace with AI developments, underscoring the urgency of the situation
2
.Singapore's Major Lenders Implement Strict Cybersecurity Controls
DBS Group CEO Tan Su Shan articulated the dual-edged nature of these advanced AI models during a briefing after the bank posted first-quarter earnings that beat expectations. "What Mythos does ... is it amplifies the risk," she told reporters, explaining that from both a speed perspective and volume perspective, the blast radius is fast. However, Tan noted that while attackers could use such tools to find weaknesses faster, banks could also use them to defend faster, with DBS still viewing AI as a net positive due to gains in coding and operations
2
.OCBC, Singapore's second-largest lender, emphasized its commitment to responsible AI use after implementing strict checks. "All such solutions undergo rigorous assessment and validation before deployment," stated Praveen Raina, OCBC's head of group operations and technology. UOB, the city-state's third-largest lender, said that while AI was a strategic pillar in its digital transformation journey, its use is governed by existing cybersecurity controls and clear internal guardrails, with a spokesperson emphasizing a disciplined and responsible approach to innovation
2
.Related Stories
Risk Amplification and the Evolving Threat Landscape
Standard Chartered Chief Executive Bill Winters characterized Anthropic's Mythos as a "sensational representation" of a broader trend of rising cyber risks, though he expressed confidence that the bank is well prepared. Winters noted that while the threat has been present for many years, the level of sophistication has increased. This perspective reflects a growing recognition among financial institutions that complex cyber threats from frontier AI represent not an entirely new category of risk, but rather an acceleration and amplification of existing vulnerabilities in software security.
The implications for the financial sector are significant. As these advanced AI models become more accessible, the window between vulnerability discovery and exploitation could narrow dramatically. Banks must now balance the operational benefits of AI—including improved coding efficiency and streamlined operations—against the heightened risk that the same technology could be weaponized by malicious actors. The intensifying checks on artificial intelligence tools across Asia's banking sector signal a shift toward more proactive risk management, with institutions recognizing that traditional security measures may no longer suffice against AI-enhanced threats. Data privacy concerns also loom large as these systems process vast amounts of sensitive information to identify potential weaknesses.
References
Summarized by
Navi
[2]
Related Stories
Banks face dual AI risk as Mythos exposes vulnerabilities while fraud attacks surge globally
20 Apr 2026•Policy and Regulation
US Officials Warn Banks About Anthropic Mythos After AI Model Finds Thousands of Vulnerabilities
10 Apr 2026•Policy and Regulation
Anthropic restricts Mythos AI model release, citing unprecedented cybersecurity capabilities
07 Apr 2026•Technology
Recent Highlights
1
Taylor Swift files trademark applications to protect voice and image from AI deepfakes
Entertainment and Society
2
OpenAI Faces Lawsuits After Failing to Report Violent ChatGPT User Before Deadly School Shooting
Policy and Regulation
3
AI chatbots provide detailed biological weapons instructions, raising urgent biosecurity concerns
Technology
Recent Highlights
Today's Top Stories
Don’t drown in AI news. We cut through the noise - filtering, ranking and summarizing the most important AI news, breakthroughs and research daily. Spend less time searching for the latest in AI and get straight to action.
