Shadow AI Poses Growing Security Risks as Enterprise Adoption Outpaces Governance

A new report warns that 'shadow AI' is spreading across enterprises as employees use public chatbots for work tasks without official approval. Business leaders say this unauthorized use is a major worry, especially as nearly half of US adults now use AI chatbots, according to a Pew Research Center survey published in March 2025.

The Pew survey found that 49% of US adults have used chatbots, up from 33% two years ago. Yet only 16% of Americans believe AI will benefit society overall. Business travelers are among the most eager to use generative AI for trip preparation, often turning to public tools rather than sanctioned enterprise systems, according to the TechRadar Pro report based on a survey by SAP Concur.

Unofficial AI usage creates compliance gaps

Shadow AI occurs when employees adopt unapproved tools for tasks such as drafting emails, summarizing documents, or analyzing data. This bypasses corporate security policies and can expose sensitive information to third-party model providers. The report notes that many workers feel forced to use public chatbots because their employers have not rolled out approved alternatives quickly enough.

• The SAP Concur survey found that nearly 60% of business travelers have used generative AI for work trips without employer knowledge.
• Yann LeCun, chief AI scientist at Meta, recently warned that AI labs like OpenAI and Anthropic face a 'big bubble explosion' because operating costs are not dropping fast enough and investor subsidies cannot last.
• Salesforce is pushing its 'agentic enterprise' vision, offering a headless 360 approach to embed AI agents across industries, but critics argue this accelerates the governance challenge.

Public trust lags behind adoption rates

Despite surging usage, public skepticism remains high. The Pew data shows that only 16% of adults think AI will benefit society, while 52% are more worried than excited. Older adults (50 and older) are the least likely to use chatbots, with 51% of all adults still not using them at all. The trust gap is especially acute in regulated sectors such as healthcare and finance, where shadow AI could trigger compliance violations under GDPR, HIPAA, or PCI DSS.

Meanwhile, debates over AI safety are intensifying. The Verge's Decoder podcast recently explored who decides when AI becomes too dangerous, highlighting Anthropic's internal evaluations of its models and the Trump administration's interest in rolling back certain safety protocols. Industry leaders say that without clearer governance frameworks, shadow AI will continue to undermine enterprise security and erode public confidence.