Risks Increase Dramatically with Employees Using Shadow AI

• AI tools based on large language models are increasingly used by staff as a quick fix to challenges in their workflow.
• If the AI tools that employees need (or want) aren’t available within the organization, users may adopt unapproved and unmanaged tools.
• This shadow AI adoption creates potential information, compliance, legal, and security risks, which are all worsened by the data-intensive nature of AI tools.
• Organizations should proactively focus on clearly communicated policies, even if informal at first, followed by data protection and AI governance, to mitigate these risks.

Regardless of role, employees face increasing pressure to complete tasks, often leading them to employ AI tools to help them do so. This issue is likely to occur regardless of which AI tools their employer officially approves.

Due to the low barrier of adoption for most web-based large language models (LLMs), users have a broad range of tools at their disposal today. If the tools they believe they need to get their work done aren’t available or have been blocked, both highly technical and novice end users are likely to find their way to the tools they want to use. This can result in users exfiltrating organizational data, creating a significant web of risk that the organization may not be able to easily untangle.