The most urgent security risks for organizations using genAI apps are data-related, and users are the critical actors in causing and preventing those risks. There are two angles to these risks.
- Input: What data do users send into genAI apps?
- Output: How do users leverage the outputs they receive from genAI apps?
For the input, the primary risk is data leakage. For the output, the risks include correctness (genAI apps are very good at providing hallucinations and misinformation) and legal concerns (many companies trained their genAI apps on copyrighted or licensed content). This report focuses on the inputs since protecting sensitive data is typically the highest priority for most organizations.
Conservative adoption
The banking industry has succeeded in slowing GenAI adoption by adopting more aggressive policies to restrict its use. 93% of banking organizations block at least one genAI app, whereas the worldwide average is only 77%. Furthermore, banking organizations blocking genAI apps tend to block more genAI apps than their counterparts in other industries. The following figure shows that the average number of genAI apps that a banking organization blocks has increased from 5 a year ago to almost 9 today, with the top 25% of organizations blocking at least 24 apps. In contrast, the global average is only 2.6 apps, with 12 apps in the top 25% of organizations.
The apps with the most blocks include apps from multiple categories, including writing assistants, chatbots, image generators, and audio generators, bearing many similarities with global trends.
On the other hand, the most popular genAI apps in the banking industry include chatbots, writing assistants, copilots, and note-taking apps, mirroring many global trends.
Organizations in the banking industry control access to allowed genAI apps using a combination of controls, including data loss prevention (DLP) and real-time user coaching. DLP as a genAI control is popular in the banking industry, with more than half of all organizations using it to restrict sensitive information from flowing into genAI apps, compared to a 43% global average. In banking, a highly regulated industry, the most common type of sensitive data uploaded to genAI apps is unsurprisingly regulated data, where DLP controls help prevent such data from being disclosed to third-party genAI vendors.
The genAI controls used in the banking industry have been very effective. The industry sees generally lower genAI adoption than other industries, with 87% of organizations using genAI compared to the global average of 97%. Although adoption continues to increase, it is doing so modestly, rising 7 points in the past year. Similarly, the average number of genAI apps used in each organization, 6, is lower than the global average of 9.6.