We Reviewed How Companies are Accessing AI. Here’s What We Found.

I’ll state the obvious: artificial intelligence is taking over email as the most transformative productivity tool to hit corporate workstations, and its reliance is growing by the day.

Its benefits like greater efficiency and improved output are outpaced only by its risks. Data breaches due to shadow AI usage are a chief concern among data security professionals, and it’s understandable why. Without the correct security frameworks and tools in place, it’s difficult to get visibility and control over sensitive data being shared with risky AI applications or large language models.

We reviewed anonymized data on how our web security customers access AI on a weekly basis to better understand not only how the average organization interacts with AI, but what security controls they have in place to mitigate its risks.

In this exercise, we segmented out company sizes based on the following user counts to help us better classify and identify trends:

• Medium-sized business: 101-999 users
• Large business: 1,000-4,999 users
• Mid-level enterprise: 5,000-9,999 users
• Large-scale enterprise: 10,000+ users 

In reviewing this data, there were a few findings that stood out to us, and we felt were worth sharing.

Large-Scale Enterprises are Embracing AI

It’s unsurprising to see that companies with more resources, departments and users are accessing more AI applications than any other business. But what is striking is the scale with which AI adoption at larger organizations has outpaced smaller ones.

Large-scale enterprises access almost five times as many AI applications per week than the average medium-sized business – and more than twice as many AI applications as the average large business.

Where you would think there might be departmental crossover between, let’s say, the finance and marketing teams both using ChatGPT, the data shows something different. These larger businesses are accessing even more AI applications to keep the wheels moving and perhaps even automate certain aspects.

Mid-Level Enterprises are Clamping Down on Shadow AI

Web security policies help organizations ensure that users are only accessing approved AI applications by blocking access to unapproved AI applications, greatly reducing the risk of shadow AI. 

A high block rate is a sign of an effective web security strategy. But it could also signal that the processes and frameworks in place for adopting AI aren’t living up to their full potential, prompting some users to ignore the rules – and companies need to be able to understand user trends, validate the risk of potential new resources balanced against the potential benefit to the business, and implement the controls necessary to facilitate secure use of resources that can promise additional benefits.

Web security policies need to strike the right balance between blocking web hits for AI apps due to disregard for existing AI usage policies – and blocking web hits for AI apps due to a lack of understanding. Even if your web security policies are tuned perfectly, if the user base doesn’t understand how to navigate the web in accordance with acceptable use policies, then you will see higher blocking rates.

The Most Popular AI Apps Have Some Unfamiliar Faces

ChatGPT, Copilot and Gemini dominate headlines, but we found our customers visit over 1,400 different AI applications per week – far more than a handful of household names.

The top 10 most popular AI applications have many familiar names but when you begin to break down AI application usage by category, it’s clear that employees are comfortable using and sharing data with a wide variety of vendors.

The Top 10 Most-Used AI Applications

• ChatGPT
• Copilot
• Codeium
• AnyClip
• QuillBot
• LanguageTool
• Perplexity
• Google Gemini
• Sapling
• Yellow

Top 5 AI Apps by Category

Chat and Search

• ChatGPT
• Copilot
• Perplexity
• Gemini
• Claude

Coding

• ChatGPT
• Copilot
• Codeium
• Gemini
• Claude

Image and Video

• AnyClip
• Synthesia
• Napkin
• Freepik
• Icons8

Writing and Grammar

• QuillBot
• LanguageTool
• ProWritingAid
• Wordtune
• Writer

Get Visibility and Control Over GenAI Usage

Forcepoint customers can use a mix of solutions to get visibility over which AI applications users are accessing, and control over the data being used within them.

Identify which AI apps web users are visiting and block or guide them to approved sites with Forcepoint Web Security (SWG).

Detect which users are sharing data within apps like ChatGPT and assess the sensitivity of that data with Forcepoint Data Security Posture Management (DSPM).

Prioritize risk and remediate data security incidents within AI applications with Forcepoint Data Loss Prevention (DLP).
Talk to an expert today to schedule a demo of Forcepoint SWG, DSPM or DLP.