Many financial institutions still rely on aging technology. This outdated reliance slows down their operations and increases exposure to insider threats, but also prevents them from using AI tools to strengthen security and reduce risk. The result is growing financial and reputational costs for any institution lagging behind.
The staggering cost of internal breaches
Insider attacks represent the most expensive breach vector across industries, averaging $4.92 million in direct costs per breach according to IBM’s 2025 Data Breach Report. For financial services firms specifically, the damage is even more significant at $5.56 million per breach.
Yet these figures tell only part of the story. Beyond the direct costs of forensic investigations, remediation and regulatory fines, insider breaches can cause severe reputational damage. Just a single incident can erode customer trust, drive away customers and permanently weaken a financial institution’s market position. For an early-stage challenger bank, these consequences could be fatal.
In the age of AI, the threat landscape has fundamentally shifted. <a href="https://newsroom.ibm.com/2026-07-29-ibm-study-one-in-four-malicious-breaches-are-ai-enabled,-costing-companies-6-million-on-average” rel=”nofollow noopener” target=”_blank”>A 2026 IBM study found that one in four malicious breaches are now AI-enabled, a 56% year-over-year increase.
Historically, finding and exploiting vulnerabilities manually was a slow process, but AI has accelerated the threat timeline. Both insiders and external attackers can now deploy autonomous agents that discover and bypass security controls at machine speed.
AI agent autonomy – even when the agent is well intentioned and activated through proper corporate protocols – compounds this risk. As McKinsey points out, AI agents can autonomously decide which data and environments to access to accomplish their tasks. Such autonomy can boost productivity, but it also heightens risk if an agent’s actions run afoul of enterprise risk controls.
When these AI agents operate without human oversight, malicious actions can easily go unnoticed for extended periods.
Insider threats persist because legacy systems let them
Siloed data creates blind spots. Legacy architectures create data silos across fragmented systems, APIs, and databases. Manual audit trails provide only retroactive visibility. The core problem here is that you cannot mitigate what you can’t see. Insider activities can remain hidden for months within these fragmented systems, causing reputational damage as long as they go undetected.
Legacy systems force dangerous workarounds. Financial institutions face a tough choice. They can deploy AI agents directly on legacy systems and risk bypassing important governance controls, creating security and reputational concerns or they can avoid deploying AI and miss out on its productivity and threat detection benefits. IBM data shows this paralysis is costly. Shadow AI now accounts for 43% of security incidents, more than double the 20% last year. Each workaround increases risk and undermines trust.
No real-time monitoring. Effective defense requires real-time behavioral monitoring to detect when AI agents or insiders act outside expected patterns, permissions or workflows. Legacy systems from the 1980s simply weren’t designed for continuous monitoring at machine speed.
Modernization is a security imperative
First, financial institutions need to integrate their data architectures to break down silos and gain a single, clear view of their operations. Second, they must orchestrate systems through governed interfaces, audit trails and robust access controls.
Financial institutions must also fight AI with AI. The speed and scale of AI-driven attacks have made humans the bottleneck in detection and response. Banks can no longer rely solely on manual remediation. If AI is being used to identify vulnerabilities, AI must also be used to help detect, contain and remediate them.
Finally, AI agents themselves must be secured. They should be treated like employees and granted only access to systems and data required to perform their tasks. Structured workflows should act as guardrails to ensure predictable behavior and reduce hallucinations or unexpected actions. Financial institutions should aim for traceable, auditable AI systems that can be monitored, governed and trusted.
Financial institutions have a choice: modernize their infrastructure or continue to face higher insider threat risks, slower breach detection and rising financial and reputational costs. With every incident, trust becomes harder to rebuild. Modernization is the only long-term solution and anything else is only a temporary fix.
Liam Leahy is CEO of UST FinX, the banking and financial services platform of UST, a leading AI and technology transformation solutions company.
