The AI-Powered Cybercrime Era

Are our financial institutions ready?

Artificial intelligence is transforming the way we work, communicate, conduct business and manage financial transactions. Its potential is enormous, but so is its potential for misuse.

The same technology that enables organizations to identify vulnerabilities, automate processes and strengthen security is increasingly being used to develop sophisticated cyberattacks. Criminals no longer need to possess exceptional technical expertise to exploit weaknesses in digital infrastructure. AI is lowering the barriers to entry, accelerating the identification of vulnerabilities and making cybercrime more accessible, scalable and difficult to detect.

October 2026 offers a particularly striking illustration of this changing threat landscape. The incidents and figures reported during the month paint a concerning picture: more than 80,000 disclosed vulnerabilities, thousands of critical security weaknesses, major infrastructure exposures, an alleged $6.3 billion cryptocurrency breach, and financial institutions introducing restrictions on third-party AI agents accessing customer accounts.

One of the most concerning developments is the reported increase in software vulnerabilities. According to the figures compiled for 2026, more than 80,000 Common Vulnerabilities and Exposures (CVEs) have been documented, compared with approximately 40,000 during 2025. More than 9,000 of these vulnerabilities were classified as critical.1

This represents a significant challenge for organizations responsible for maintaining secure digital infrastructure. However, the number of vulnerabilities alone does not tell the whole story. What makes the situation particularly alarming is the speed at which cybercriminals can turn a newly discovered weakness into an actual attack.

Historical figures illustrate this changing reality. The reported time between vulnerability disclosure and exploitation has fallen from approximately 756 days in 2018 to 32 days in 2022 and just five days in 2023.2 By 2026, the concern is that attackers may exploit certain weaknesses before organizations have even deployed the relevant security patches. 

One reported statistic suggests that 26.3% of vulnerabilities are exploited within the first 24 hours. This fundamentally changes the responsibilities of cybersecurity teams. Traditional approaches, where organizations identify weaknesses and address them through periodic security updates, may no longer be sufficient. Security must become continuous, proactive and capable of responding at the same speed as emerging threats.

The reported PHP-to-shell vulnerability provides an important example of the potential consequences of weaknesses embedded in widely used technologies. PHP supports a substantial proportion of websites worldwide, including platforms built on WordPress. According to the compiled material, a vulnerability that had remained undiscovered for approximately 14 years was identified through AI-assisted security research.3

AI will continue to create opportunities for innovation. It will also create opportunities for criminals. The difference will increasingly depend on how responsibly organizations deploy technology, govern access and protect the people and systems entrusted with their customers' assets.

The potential exposure was considerable, with an estimated 22.6 million PHP instances affected.4 The incident illustrates a difficult reality: a vulnerability does not need to be newly introduced to become dangerous. A weakness that has existed for years can suddenly become a major security concern once discovered and understood by malicious actors.

The reported attacks involving PHP infrastructure, followed by the identification of another code execution weakness,5 demonstrate why organizations cannot afford to assume that established technologies are automatically secure.

Similar concerns were raised regarding Node.js and vulnerabilities affecting other widely used technology platforms. For organizations, particularly banks and financial institutions, the implications extend beyond website security. Their digital infrastructure includes mobile applications, APIs, third-party platforms, payment gateways and interconnected systems. A weakness in any one of these components may create an entry point into a much larger environment.

Artificial intelligence has also changed the economics of vulnerability discovery. Bug bounty programmes traditionally encourage ethical hackers and security researchers to identify weaknesses in software and report them to organizations in exchange for rewards.

However, AI-assisted research has significantly increased the volume of submissions. Approximately 1,390 reports were submitted during the first half of 2026, compared with around 850 throughout 2025. 

The problem is that a greater volume of submissions does not necessarily translate into a greater number of genuine security weaknesses. When security teams are overwhelmed with reports, distinguishing genuine vulnerabilities from irrelevant or inaccurate submissions becomes increasingly difficult. The result is additional pressure on resources that are already responsible for protecting complex digital environments.

This creates an important paradox: AI is helping identify security weaknesses faster, but organizations may not have the resources to investigate and remediate them at the same speed.

Cybercrime involving digital assets presents another dimension of the growing threat. The recent Binance incident, involving an alleged $6.3 billion breach, illustrates the potentially devastating consequences of compromised credentials and excessive privileged access.

According to the incident account included in the source material, attackers allegedly compromised a DevOps engineer's laptop and exploited the associated administrative access to reach service infrastructure.8 The attackers reportedly prepared transaction signatures in advance and executed transactions across ten blockchains9 within approximately five seconds.

Of the reported $6.3 billion, approximately $3 billion was frozen,10 while $3.3 billion reportedly moved beyond the platform's custody. The incident, if confirmed in the reported form, raises questions that extend far beyond conventional cybersecurity.

How much authority should a single employee or device possess? Should one administrative account be capable of facilitating transactions involving billions? Are transaction approval mechanisms sufficiently independent? Can financial institutions and digital asset platforms detect and interrupt unusual movements before assets leave their control?

These are fundamentally questions of governance, internal controls and operational risk. For cryptocurrency businesses, security cannot depend solely on protecting passwords, private keys or individual devices. It requires a combination of segregation of duties, transaction limits, independent verification, time delays, privileged access management and continuous monitoring.

The principle is straightforward: no individual, device or compromised account should possess unrestricted authority to transfer substantial financial assets.

Perhaps one of the most important developments for the financial sector is the growing concern surrounding autonomous AI agents. AI agents differ from conventional AI applications because they can be designed to perform tasks, interact with external systems and execute actions with limited human intervention.

While this offers operational benefits, it also introduces a new category of risk. An AI agent with access to financial accounts, customer information or transaction-related services could potentially expose sensitive information or initiate actions beyond the intended scope of its authority.

Another emerging issue is Shadow AI. Shadow AI occurs when employees use AI applications without the knowledge, approval or oversight of their organizations. For example, an employee may copy customer information, account details or confidential business documents into an unauthorized AI platform to complete a task more quickly.

The consequences can be serious.

Sensitive information may leave the organization’s-controlled environment without the necessary authorisation, creating potential confidentiality, privacy, regulatory and data protection concerns.  For banks, this is not merely an IT issue. It is an information governance, operational risk and compliance issue. 

Across the incidents described in the source material, a recurring pattern emerges: sophisticated cyberattacks frequently exploit weaknesses in trusted relationships. Compromised credentials, inadequate vendor controls, unsecured infrastructure, forgotten test environments and poorly managed access privileges can provide attackers with opportunities that conventional security measures may overlook.

The reported Microsoft 365 compromise, affecting 173 tenants through a stolen third-party credential,12 is one example of the risks associated with external access. Similarly, the reported exposure of thousands of customer records and the disruption caused by compromised infrastructure highlight how incidents can extend beyond the organization initially targeted.

This is particularly relevant to financial institutions, where operations increasingly depend on technology providers, cloud services, external applications and interconnected payment systems. A bank may maintain strong internal security controls, but its overall security posture can still be compromised through a vulnerable service provider or an inadequately monitored third-party connection.

Maintaining an accurate inventory of digital assets, monitoring privileged access, reviewing third-party relationships and regularly assessing vulnerabilities should therefore be essential elements of institutional security.

Privileged accounts should operate under strict controls. Multi-factor authentication, role-based access, segregation of duties and independent approvals should be mandatory for sensitive systems and high-value transactions.

Organizations should maintain updated inventories of internet-facing systems, applications, APIs and third-party connections. Vulnerabilities must be prioritized according to their severity, exposure and potential business impact.

Financial institutions need clear policies governing the use of AI tools and autonomous agents. Vendor relationships should be assessed not only during onboarding but throughout their lifecycle. Access permissions, security standards, incident notification requirements and data protection arrangements should be regularly reviewed.

Cyber-enabled financial crime frequently overlaps with fraud, identity theft, account takeover, money laundering and unauthorized movement of funds. Financial institutions should ensure that cybersecurity incidents can trigger appropriate fraud investigations, transaction monitoring and AML escalation processes.

AI-generated phishing messages, impersonation attempts and cloned voices are making social engineering increasingly convincing. Employees and customers must understand the importance of independent verification, secure authentication and callback procedures.

Offline backups, tested recovery arrangements, clear incident response responsibilities and regular simulation exercises can significantly reduce the impact of a successful attack.

The challenge is not simply to develop stronger firewalls, deploy more sophisticated software or purchase additional security tools. It is to build organizations capable of anticipating threats, identifying weaknesses, responding quickly and limiting the damage when preventive controls fail.

AI will continue to create opportunities for innovation. It will also create opportunities for criminals. The difference will increasingly depend on how responsibly organizations deploy technology, govern access and protect the people and systems entrusted with their customers' assets.

内容来自 Pakistan Today,版权归原作者所有。

阅读原文 ↗

把喜欢的内容,留在你的新标签页。

打开我的订阅