The image of the bank robber is hopelessly outdated. Today’s heist does not begin with a getaway car outside a branch. It begins quietly, with an adversary establishing persistence inside a financial institution’s network and studying how the organisation responds, says Tom Kellermann, VP of AI Security and Threat Research at Trend AI.
The objective is no longer simply to steal money. Attackers want access to the financial system itself: payment infrastructure, confidential market intelligence, investment strategies, customer identities and the mechanisms used to defend them. TrendAI’s Modern Bank Heists in 2026 research, based on a survey of 48 financial-sector CISOs, reveals the scale of this shift. Eighty-nine per cent reported an increase in AI-enabled attacks, while 55 per cent had experienced a rise in attacks against APIs. Almost half, 46pc, had encountered attempts to steal non-public market information or investment strategies.
These are not smash-and-grab operations. They are sophisticated campaigns of economic espionage, market manipulation and systemic disruption. More concerning still, 67pc of CISOs said their institutions had experienced what I call “counter-incident response”. This is when attackers actively undermine the defenders attempting to contain them. They monitor security teams, corrupt evidence, disable tools, change tactics and target the responders themselves.
The adversary is no longer merely evading detection. It is fighting back.
Cybercriminal organisations increasingly operate like cartels. They specialise, collaborate and share access, infrastructure and expertise. One group gains entry, another establishes persistence, and others conduct fraud, extortion or data theft. AI allows them to accelerate reconnaissance, automate social engineering and adapt attacks at a speed few human-led security operations can match.
Techniques such as AI-enabled spear phishing, deepfake-assisted business email compromise and mobile phishing are making identity-based attacks more credible and scalable. Remote access tools continue to evolve, while steganography enables malicious code or stolen information to be concealed inside apparently harmless files. Financial institutions are also deeply interconnected. An attacker who compromises one bank, technology provider or trusted partner may use that position to target others. This “island-hopping” turns individual compromises into campaigns against an entire financial ecosystem.
Yet many financial institutions are attempting to confront this threat with flat budgets and outdated governance. Fifty-four per cent of the CISOs surveyed received no increase in security funding, despite the escalation in hostile activity. There is also a fundamental conflict of interest within too many institutions. Fifty-seven per cent of security functions still report to the CIO. The person responsible for protecting the institution therefore reports to the executive responsible for connectivity, digital transformation and operational availability.
Defensive co-ordinators should not report to offensive co-ordinators.
Cybersecurity must be treated as institutional defence, not merely an IT function. The CISO needs independence, authority and direct access to the board. Security teams also need the mandate to suppress intrusions at machine speed. For years, the industry has prioritised visibility. Banks have invested heavily in systems that produce more alerts, more dashboards and more information. But visibility without the authority to act simply gives an institution a better view of its own compromise.
When an AI-enabled adversary can move through an environment in seconds, waiting for layers of human approval is not prudent governance. It is an operational vulnerability. Autonomous defence does not mean surrendering control to machines. It means defining trusted parameters in advance so defensive systems can isolate compromised assets, revoke malicious sessions, disrupt lateral movement and protect critical data before an incident becomes a crisis.
Financial institutions should assume that determined adversaries will eventually penetrate their perimeter. The priority must therefore shift from attempting to prevent every intrusion to suppressing hostile activity before it achieves its objective.
That requires continuous threat hunting, stronger identity controls, comprehensive API security and greater scrutiny of trusted third parties. Institutions must also monitor for behaviour indicating that an attacker is studying or interfering with their incident-response processes.
The modern bank heist is already under way. It is persistent, automated and increasingly unconstrained by geography. Its targets extend far beyond customer accounts to the confidential knowledge and critical systems that underpin financial markets. The financial sector cannot defend itself against machine-speed attacks through human-speed decision-making. Banks must give their security leaders the independence, resources and authority to fight back before today’s hidden intrusion becomes tomorrow’s systemic event.
