Taiwanese banks turn to AI defenses as cyberattack activity climbs
New bank disclosures show heavier attack traffic and growing use of AI for detection. Giving software authority to act remains a separate decision.
Loading page…
New bank disclosures show heavier attack traffic and growing use of AI for detection. Giving software authority to act remains a separate decision.
Listen to this story
Taiwanese banks are adding AI-assisted detection to security operations as attack activity grows, while keeping consequential decisions under human review and treating autonomous agents cautiously. The operational challenge is broader than identifying AI-driven attacks: automated scanning is difficult to distinguish, and exercises have exposed gaps in filtering, application security, permissions and incident response. The banks’ disclosures point to a near-term focus on using AI to triage threats and test defenses, not handing security decisions over to agents.
Taiwan Cooperative Bank’s average monthly suspicious probes and attacks through October 2026 was about twice its 2025 average.
About 80% of the cooperative bank’s first ten months of activity occurred from August through October, targeting several digital banking services.
First Bank’s web application firewall blocked about 2.2 million attacks from January through October 2026; the figure counts blocked attempts, not breaches.
Taiwanese banks are putting AI to work against rising cyberattack traffic, but they have not confirmed that autonomous AI is behind the attacks. In disclosures published by CNA on November 22, Taiwan Cooperative Bank reported roughly double its previous year’s monthly attack activity, while banks described AI tools for spotting suspicious behavior and assessing risks.
First Bank and Taiwan Cooperative Bank said they had found no attacks they could clearly attribute to AI acting autonomously. Bank of Taiwan said attack methods were already so automated that identifying AI’s involvement was difficult. That distinction leaves the banks facing a measurable increase in hostile activity without a clear answer about how much of it involves AI.
Taiwan Cooperative Bank’s monthly average of suspicious external probes and attacks through October 2026 was about twice its 2025 average. Activity rose quarter by quarter, with roughly 80 percent of the first ten months’ total concentrated between August and October. Targets included its website, online banking, corporate banking and electronic billing services. The bank said it had successfully defended those services.
First Bank said its web application firewall, which filters traffic to web services, blocked about 2.2 million attacks. This is a count of blocked attacks, not successful breaches.
First Bank also saw probing activity increase significantly from the third quarter, mainly against public-facing websites. Bank of Taiwan described the activity it detected primarily as automated scanning and checks for security weaknesses.
The banks described different stages of adoption, from deployed detection tools to testing. First Bank said AI analysis reduces the number of monitoring events staff must inspect individually, helping its security teams monitor threats and respond more efficiently.
More autonomous systems face additional limits. First Bank is evaluating future use of AI agents for security monitoring and incident analysis, alongside safeguards. Taiwan Cooperative Bank has begun putting a small number of low-risk agents into workflows and is developing controls covering data security, permissions and human review.
Rays Technology consultant Chen Chia-hung urged banks to go beyond using AI to detect attacks after they begin. He recommended using it to simulate attacks, uncover weaknesses and test defenses before hackers exploit them. He also called for AI-driven attacks to become part of routine security exercises, according to Focus Taiwan.
Existing exercises have exposed practical gaps. First Bank said a drill involving traffic designed to overwhelm services showed that its internet provider’s traffic-filtering arrangements needed improvement; the two worked on adjustments. Taiwan Cooperative Bank said this year’s simulated attacks identified room to improve application security and account permissions, with remedial work still being tracked.
Chen said drills also need to establish whether people can execute response plans: reach network providers, get suppliers involved quickly, contact investigators and notify regulators. For consequential decisions, the banks are retaining a human checkpoint. First Bank said important judgments or actions require human review, while Bank of Taiwan said people must review significant business judgments or decisions in future agent applications.
Loading discussion...
Join the conversation
Explain where you would draw the line between speed and oversight.
Be the first to share a perspective or an experience.
Reader comments
Newest comments first. Replies stay oldest first.