What is Shadow AI? Definition, risks, and examples - LastPass

GUIDE

Shadow AI: the unauthorized AI that’s already in your stack

Only 27% of organizations have an actively enforced governance program, despite AI being in active use at 92% of organizations. See the practical framework lean IT teams can use to close the gap.

Get the free report

Trusted by 100,000+ businesses worldwide · SOC 2 Type II, ISO 27001, ISO 27701 Certified · 4.5/5 on G2

Key stats: What the data shows

Source: The LastPass 2026 State of AI and SaaS Security Report. Insights based on survey responses from 412 IT leaders, login habits from 450,784 users across 20,631 organizations, and app usage at 15,907 organizations.

92% of IT leaders say AI is in active use across their organizations.

Why it matters: LastPass survey data shows AI adoption is effectively universal, but governance continues to lag. The report explores how that gap varies by organization size and adoption maturity levels. 

46% of organizations rate their shadow AI exposure as moderate to high.

Why it matters: Nearly half are rating their shadow AI exposure as moderate to high, yet only 22% have tools to actively monitor or log AI usage. The organization's sense of its own risk isn’t matched by any mechanism to act on it.

56% of IT leaders are concerned about credentials entered into AI prompts.

Why it matters: If credentials are entered into an AI prompt during an unmonitored session and that session is compromised, an attacker can reuse those credentials and potentially pivot into connected systems.

Why shadow AI is structurally harder to govern than shadow IT

Traditional shadow IT leaves a footprint. An employee installs unapproved software, and endpoint detection flags it. Or they connect an unrecognized device, and network monitoring surfaces it.  

Shadow AI doesn't work that way. The tools with the highest adoption rates, like ChatGPT (used in 64% of organizations) and Claude (used in 39%) are browser-based. 

The only evidence produced is a login executed in the browser, using either personal or corporate credentials. 

And therein lies the challenge. Login events can blend into trusted activity and hide shadow use from scrutiny.  

This lack of visibility leads to downstream effects such as: 

  • Sensitive data entering AI tools, cited by 68% of IT leaders who told LastPass this was a top concern. 
  • Passwords and login credentials entering AI prompts, cited by 56% of IT leaders in LastPass’s survey as another major concern. 
  • AI tools accessing corporate systems, flagged by 61% of IT leaders surveyed by LastPass, who worry about unauthorized actions. 

Because shadow AI can occur within normal logins in ordinary browser sessions, AI governance must start with building visibility where access occurs. Without that visibility, your organization is less equipped to address data exposure risks and rising threats like prompt injections. 

Want to see AI governance trends among real companies in 2026? The full governance data is in the LastPass 2026 State of AI and SaaS Security Report.  

Get the complete insights

Download the report

How shadow AI discovery works and why most approaches miss it

The methods most IT teams reach for when attempting shadow AI discovery have one thing in common: They struggle to identify browser-based AI usage.

  • Employee surveys ask people to self-report which AI tools they use. The results skew toward undercounting as users don't automatically register Grammarly's AI rewrite feature or Notion AI’s Q&A capabilities as "using an AI tool."  
  • Network monitoring and endpoint management are built to detect traditional shadow IT, which typically creates discoverable technical evidence like software installs or file transfers. 
  • Blocked-tool lists work the same way, addressing tools you've already identified. They do little to discover what you haven't.

Organizations achieving better visibility often focus on capturing login activity regardless of credential type.

LastPass Business Max operates at the authentication layer, which is the only layer where browser-based AI tools are consistently visible regardless of the credentials employees use for access.

When an employee signs into ChatGPT, Claude, or Grammarly using a personal email, free-tier account, or work credentials, the authentication event is logged via the LastPass browser extension.

This is what separates LastPass from endpoint management and network monitoring: Those tools look for software installations and unusual traffic patterns, but shadow AI generates neither. It generates a login, and that login is what LastPass captures.

 
Try LastPass Business Max free Get a demo

What shadow AI visibility makes possible

Shadow AI visibility doesn't eliminate AI. But that’s not the goal. The focus is to move unmonitored adoption into a managed framework, where IT can make informed decisions about use. 

With access-layer visibility into shadow AI: 

  • You can see which AI tools employees are accessing with personal credentials, the access pattern that 56% of IT leaders told LastPass they’re concerned about but can’t currently track. 
  • You can apply tiered controls based on your organization’s security obligations or where risks justify intervention. 
  • Shadow AI stops growing silently. Employees who encounter in-flow warnings in the browser understand that the organization has visibility, which may inspire more caution and care about which apps they access. 
  • You can demonstrate, when auditors or insurers ask, that controls over AI tool access exist and that your inventory reflects real usage, not self-reported estimates. 

The starting point is always the same: discovering which AI tools your employees are using. 

The full data on AI app usage and governance gaps

The LastPass 2026 State of AI and SaaS Security Report combines behavioral telemetry with survey data. It shows how credential exposure varies by company size and what the AI governance gap looks like from the inside. 

Download the LastPass 2026 State of AI and SaaS Security Report

Download the report

Explore more in the AI & SaaS Security series

GUIDE

AI Governance Compliance: Policies, Controls, and Gaps

Read the guide

View all AI & SaaS security guides

FAQ: What is shadow AI?

The three most-cited shadow AI security risks in the LastPass 2026 State of AI and SaaS Security Report are: 

  • Sensitive data entering AI tools without oversight, cited by 68% of respondents as their top concern. 
  • AI tools accessing corporate systems without authorization, flagged by 61%. 
  • Credentials being entered directly into AI prompts, a concern for 56%. 

Often, these interactions happen in the browser, and there are no access controls to enforce data handling requirements after the fact. 

Shadow AI and unsanctioned AI are closely related but not identical.

Unsanctioned AI refers to tools that haven’t been formally approved by IT. The organization knows the tool category exists but hasn't authorized its use.  

In contrast, shadow AI is both unapproved and unknown. It refers to usage by employees: 

  • Accessing free consumer AI tools with personal emails 
  • Sharing corporate data with AI chatbots via APIs or MCP connectors 
  • Using AI capabilities within existing, previously approved SaaS apps 

The LastPass 2026 State of AI and SaaS Security Report found that 46% of organizations rate their shadow AI exposure as moderate to high, but only 22% have any tools to govern AI usage. 

That’s where LastPass plans like Business Max provide the most value, allowing you to get visibility into non-sanctioned AI tools in use and govern that usage by applying allow, warn, or block rules. 

Discovering shadow AI is a critical early step in AI governance. Once you have visibility, you can assess each tool for security, regulatory, and data protection risks. 

Based on that assessment, you can apply appropriate controls.

For small to mid-sized teams, LastPass makes AI governance practical and easy to adopt: 

LastPass Business Max shows you what’s happening. It gives you visibility into what AI tools, SaaS apps, and websites employees are accessing, whether that’s with personal or corporate credentials.

Beyond visibility, LastPass also lets you control what’s happening. With Business Max, you can: 

  • Notify users that they’re attempting to access a questionable tool. 
  • Block entire categories of high-risk sites at once rather than app by app. 
  • Use the AI usage guidance feature to caution employees who try to share sensitive data with an AI tool. 

These controls let you govern AI use with enforcement that matches the risk level of the app or site. They’re a practical way to catch shadow AI usage and manage it without adding enterprise-grade complexity or overhead. 

Don't see your questions here? Visit Support Center.

Inizia a usare un piano LastPass per aziende

Prova LastPass gratis Contatta le vendite
Non è necessaria una carta di credito per la prova gratuita di 14 giorni.