AI is being adopted at breakneck speed within the financial services industry, introducing a new threat: shadow AI. 

Shadow AI is the unauthorized use of AI by an employee, Derek Holt, chief executive of IT solutions provider Digital.ai, told Bank Automation News.

“Large enterprises and FIs make rules on what tech solutions [like AI] employees are allowed to use or not,” Holt said. “In many organizations, the first reaction for AI was: ‘We don’t allow those websites to be accessed from our work computers.’” 

Nearly 63% of developers are using gen AI tools that are not approved by their company, Holt said, referencing a July 2024 Capgemini report. Unauthorized use of AI tools comes with multiple risks for businesses, he said, including: 

• Security concerns; 

• Legality of usage; and  

• Inability to maintain quality control. 

Eliminating gray zone 

As AI usage becomes entrenched, financial institutions are looking to slowly eliminate the use of AI outside of company purview, Rohit Bhat, managing director of financial services and capital markets at Google Cloud, told BAN. 

“People are building sidebar projects and have these technologies assist them in their daily work roles, tasks and routines, so that is definitely going to be a problem,” he said.  

Google Cloud is helping businesses bring AI tools into their operations and add custom security protocols and usage directories to provide employees with the tools in a safe manner, Bhat said. This avoids employees going outside of their companies’ secure IT environments to access unverified tools, he added. 

Google Cloud banking clients are rapidly deploying AI to shut down “unwarranted use of AI” by employees, he said. 

The tech giant’s Vertex AI platform allows FIs to bring in AI models from competitors, like Meta’s Llama models and Mistral AI, so that enterprises have access to multiple AI models and are not locked into using just one, Bhat said. 

Some institutions opting for a multiple model AI approach from AI providers like OpenAI, Google Gemini and Amazon Web Service’s Bedrock are: 

• Nasdaq; 

• Morgan Stanley; and 

• NatWest. 

“We knew that AI can be a great addition to our workflows in making us more efficient. … Our employees also wanted to use the tool,” Jason Bressler, chief technology officer at wholesale mortgage seller United Wholesale Mortgage previously told BAN.  

The lender is using Google Cloud’s Gemini AI model for document extraction and information synthesis to aid underwriters, Bressler said. It has 99% accuracy and has helped its underwriters increase their output by nearly 100%, he added. 

“Rather than the experimenting with unsafe tools, we teamed up with Google Cloud to use the technology in a controlled manner.”