0
Skip to content

Pennsylvania Bank Addresses Shadow AI Breach After Employee Uploads Customer Data

Blue tech image with a lockA bank employee exposed sensitive customer information by uploading it to an unauthorized artificial intelligence application, underscoring the risks of unsanctioned AI use.

06/15/2026 12:00 P.M.

A Pennsylvania community bank recently suffered a data exposure incident after an employee input nonpublic consumer information into an unapproved generative artificial intelligence application. The uploaded data included highly sensitive personal records, such as names, Social Security numbers, and dates of birth.

Bank representatives confirmed they had successfully contacted the vendor of the application to have the customer data removed before the vendor’s models could use it for training, according to an American Banker report.

The bank voluntarily filed a disclosure with the Securities and Exchange Commission on May 7, 2026.

“The same employee mistake triggered three separate obligations at once: SEC disclosure, a 36-hour notice to the bank’s prudential regulator and customer notice under the Gramm-Leach-Bliley Act,” according to American Banker.

The incident brings renewed attention to the operational vulnerabilities created by shadow AI — the workplace use of unauthorized or unmonitored artificial intelligence tools.

In a LinkedIn post, Jesse Kerr, an AI governance specialist, said that this incident represents a breakdown in corporate visibility rather than a traditional security perimeter failure. The bank experienced no external network compromise or stolen employee credentials. Instead, the data exposure occurred entirely because an authorized employee interacted with an unauthorized tool, leaving the organization with no real-time oversight of where the information was sent.

ACA’s Take

This breach highlights the growing risk of shadow AI in the workplace. Research from WalkMe found that 78% of employees use AI tools that are not provided or approved by their employers, ACA reported. Additionally, a McKinsey & Company report noted that corporate leadership drastically underestimates this behavior, believing only 4% of staff use generative AI regularly for a third of their work, compared to an actual employee self-report rate of 12%.

The financial consequences of unmonitored system use are also climbing. According to IBM’s 2025 Cost of a Data Breach Report, shadow AI has become one of the top three costliest factors in data breaches. The report found that 63% of breached organizations lacked any formal AI governance policies, and 97% lacked proper access controls on their active AI systems.

“I don’t know of any employee that would intentionally, maliciously try to download risky AI tools,” Heath Morgan, a partner at Martin Golden Lyons Watts Morgan, noted during an ACA Huddle. “They’re doing this to bring efficiencies to your company, and you can appreciate that initiative, but also give them a framework of how they can adopt and use new technology.”

That framework involves clear rules governing how employees can use AI and training them on the differences between secure corporate software and public, open-source applications.

“Most compliance teams can name the AI tools they sanctioned, but they have almost no visibility into what employees connected to their corporate systems,” Ofer Klein, CEO of Reco, an AI and agent security platform provider, told Forbes. “And that’s the problem, because every CISO I talk to wants to say yes. They want the business to move at the speed of AI. You can’t do that when you don’t know what’s already running.”

Related Content from ACA:

Remember, subscribe to ACA Daily and Member Alerts under your My ACA profile when logged in to acainternational.org.

If you have executive leadership updates or other member news to share with ACA, contact our communications department at comm@acainternational.org. View our publications page for more information and our news submission guidelines here.

One moment please...

Share Profile

This site utilizes cookies as included in our Privacy Policy and User Agreement.