As businesses become increasingly cautious about deploying artificial intelligence, their employees appear to be heading in the opposite direction.

New research from UK technology services provider Acora reveals that Google searches related to hiding AI usage have surged by 489% over the past two years, raising questions about whether corporate restrictions are inadvertently encouraging employees to use AI tools without their employers’ knowledge.

The findings highlight a growing challenge for businesses attempting to balance the productivity benefits of generative AI with concerns about data privacy, security and regulatory compliance.

While corporate leadership teams debate the risks of AI adoption, employees are increasingly exploring ways to use the technology independently, potentially creating a new generation of cybersecurity vulnerabilities.

The Growing Shadow AI Problem

Acora analysed four years of UK Google search data across 50 search terms associated with potentially unauthorised AI activity.

These included queries about deleting ChatGPT conversations, bypassing workplace restrictions, concealing AI-generated content and determining whether popular AI tools are safe for confidential information.

The research identified 161,934 UK searches related to potentially unsanctioned AI use over the past year, representing a 48% increase compared with two years earlier.

That figure is also approximately 2.6 times the search volume recorded during ChatGPT’s first year.

Perhaps more revealing is the dramatic growth in searches specifically associated with concealing AI activity.

Queries about deleting or hiding AI chat histories increased from 1,861 to 10,958 over two years, a rise of 489%.

Three categories of potentially risky behaviour — concerns about data safety, efforts to conceal AI activity and attempts to bypass restrictions — collectively accounted for more than 100,000 searches over the past year.

This represents more than three times the level recorded two years earlier.

The findings suggest that interest in AI tools is continuing to accelerate even as organisations reconsider how quickly they should deploy the technology.

This growing disconnect reflects the broader challenge of shadow AI, where employees adopt artificial intelligence tools outside officially approved corporate systems.

As Tech Trends recently explored in its coverage of NordLayer AI Scout and the growing security risks of shadow AI, organisations are increasingly struggling to understand which AI applications employees are using, what information they are sharing and where sensitive corporate data might be going.

However, Acora emphasises that search volumes indicate interest and possible intent rather than confirmed workplace behaviour. The data cannot establish whether individual searches were conducted by employees or whether AI tools were subsequently used in violation of company policies.

Employees Are Looking for Ways to Cover Their Tracks

One of the most striking findings concerns the growing interest in removing evidence of AI activity.

Searches for “how to delete ChatGPT history” reached 4,052 over the past year, an increase of 418% compared with two years earlier.

These trends raise questions about whether employees are becoming increasingly concerned about how their AI usage might be monitored.

Of course, deleting chat histories is not necessarily evidence of wrongdoing. Many users routinely clear their browsing histories and conversations for legitimate privacy reasons.

Nevertheless, the growing interest in these activities presents challenges for organisations seeking visibility into how AI systems interact with corporate information.

When employees use unapproved AI services, sensitive documents, customer information or proprietary business data may be processed outside established security controls.

If those interactions are subsequently deleted or otherwise concealed, businesses may struggle to establish what information was shared and whether any data protection obligations have been compromised.

Confidential Data Concerns Are Rising Rapidly

Acora’s research also reveals growing uncertainty about whether AI platforms can safely handle confidential business information.

Searches for “is ChatGPT safe for confidential information” increased by 1,046% over two years, rising from 144 to 1,650.

Questions about Microsoft’s Copilot tools have also increased significantly.

Searches for “is Copilot safe to use at work” increased by 159% over the past year.

These figures suggest that users are increasingly aware of potential privacy and security implications but may lack clear guidance about what information can safely be shared with AI systems.

The distinction between consumer AI services and enterprise-managed deployments is particularly important.

Enterprise AI platforms may offer additional administrative controls, contractual data protections and information governance capabilities. However, those safeguards depend on the specific product, configuration and organisational policies.

Simply approving an AI tool does not automatically guarantee that employees understand how to use it securely.

The issue also reflects a wider challenge surrounding AI-enabled technologies and personal information.

For example, a recent Tech Trends investigation into the hidden privacy risks of AI-powered smart glasses highlighted how emerging AI devices can collect and process sensitive information through cloud-based services, sometimes with limited transparency about data retention and third-party access.

Although workplace AI assistants and smart glasses serve different purposes, both illustrate the importance of understanding where information is processed, who can access it and how long it remains available.

For businesses, those questions become particularly important when employees use AI tools without consulting IT or security teams.

Businesses Are Slowing Down While Employees Push Ahead

The growing interest in potentially unauthorised AI activity comes at a time when many organisations are reconsidering their AI deployment strategies.

According to research from FTI Consulting, 60% of businesses have slowed, paused or scaled back planned AI deployments because of reputational, regulatory or trust concerns.

This creates an increasingly difficult situation for IT leaders.

Organisations may be delaying AI deployments to ensure appropriate safeguards are established, while employees continue looking for tools that can help them complete everyday tasks more efficiently.

Writing reports, summarising meetings, analysing documents and preparing presentations are among the activities where generative AI can offer immediate productivity benefits.

When approved solutions are unavailable, employees may be tempted to turn to free consumer alternatives.

The resulting phenomenon, commonly known as shadow AI, presents many of the same challenges as shadow IT, but potentially with greater consequences for information security.

Unlike traditional unauthorised software, generative AI applications can receive and process substantial amounts of sensitive information through ordinary text prompts.

Employees may inadvertently expose confidential business information simply by asking an AI assistant to summarise an internal document or analyse customer data.

The challenge is therefore not limited to identifying which applications employees access. Organisations must also understand how those applications interact with corporate information.

AI Restrictions May Be Creating New Risks

Another important finding concerns attempts to bypass restrictions on AI services.

Acora’s research identified a substantial increase in searches associated with circumventing AI blocks, suggesting growing interest in accessing services that may be unavailable through conventional channels.

Although some of this activity may originate from students attempting to access AI tools on restricted educational networks, the trend raises broader questions about the effectiveness of blocking access to AI platforms.

Employees determined to use generative AI may simply move to alternative services, including platforms that require no registration.

Such services can present additional governance challenges because organisations may have limited visibility into their data retention policies, security practices or the information being processed.

AI-powered meeting assistants represent another potential area of concern.

If employees use unapproved meeting assistants to record confidential conversations, businesses could inadvertently expose sensitive information to third-party services.

This becomes particularly problematic when discussions involve customer data, commercial negotiations, financial information or personnel matters.

The situation becomes even more complex as organisations begin adopting autonomous AI agents capable of interacting with multiple applications and performing tasks independently.

As examined in Tech Trends’ recent expert analysis, monitoring an AI agent is not the same as being able to stop it.

While visibility into AI activity is essential, businesses also need mechanisms to revoke permissions, halt delegated tasks and independently verify that automated systems remain within their authorised boundaries.

For organisations already struggling to track employee use of conventional AI chatbots, the emergence of increasingly autonomous systems adds another layer of complexity.

The Rise of AI Content Disguising Tools

Beyond hiding conversations and bypassing restrictions, Acora also examined the growing popularity of tools designed to make AI-generated writing appear more human.

These so-called AI humanisers are increasingly marketed as ways to rewrite machine-generated content so that it is less likely to be identified by AI detection systems.

Although much of the demand may originate from students, the popularity of these tools illustrates another challenge for organisations attempting to establish transparency around AI-generated work.

If employees use AI to produce reports, presentations or other business materials without disclosure, managers may find it difficult to assess whether the information has been independently verified.

The issue is not necessarily the use of AI itself, but the absence of appropriate review and accountability.

AI-generated content can contain factual inaccuracies, misleading references or confidential information that should not have been included.

Disguising its origins may make those problems harder to identify.

However, AI detection tools themselves are not consistently reliable, meaning that attempts to distinguish human-written from machine-generated content can also produce misleading results.

For businesses, establishing clear expectations around disclosure, verification and accountability may therefore be more effective than relying exclusively on automated detection.

Why AI Governance Needs More Than Restrictions

For businesses, the findings suggest that traditional approaches to technology governance may be insufficient.

Simply prohibiting certain AI tools does not eliminate employee demand for the technology.

Instead, restrictive policies without practical alternatives may encourage employees to seek less visible solutions.

Acora argues that organisations should use periods of slower AI deployment to improve their understanding of how employees are already using the technology.

This includes identifying AI applications accessed through corporate networks, reviewing browser extensions and examining AI capabilities embedded within existing software.

Businesses also need clearer policies explaining which information employees can share with AI systems and which tools are approved for particular tasks.

Training remains an important consideration.

Employees may understand that certain information is confidential without appreciating how generative AI systems process prompts, retain conversation histories or interact with external services.

Providing practical guidance about acceptable AI use can help reduce accidental data exposure while enabling employees to benefit from the technology.

Businesses should also consider whether existing approved tools adequately meet employee needs.

If staff consistently seek alternative AI services, this may indicate gaps in functionality, accessibility or usability rather than deliberate disregard for security policies.

Providing employees with secure, effective alternatives to consumer AI tools could reduce incentives to circumvent restrictions.

For example, enterprise AI deployments can be configured with appropriate access controls and data governance measures, helping organisations support productivity while maintaining oversight.

However, these safeguards require ongoing management as AI platforms introduce new features and capabilities.

Finding the Balance Between Innovation and Security

The broader implications of Acora’s findings extend beyond the UK.

As generative AI becomes increasingly integrated into everyday professional workflows, organisations worldwide face similar questions about how to encourage innovation without compromising security.

The challenge is particularly significant because AI adoption is no longer confined to dedicated technology teams.

Employees across marketing, finance, human resources, legal services and customer support can now access powerful AI capabilities through ordinary web browsers.

That accessibility makes AI difficult to govern using conventional software approval processes alone.

The findings also highlight an important distinction between having access to artificial intelligence and being adequately prepared to use it.

As Tech Trends recently reported in its coverage of the UK’s position in a Commonwealth AI readiness study, technological infrastructure alone is insufficient to ensure that individuals and organisations can realise the benefits of AI.

Skills development, responsible adoption and appropriate governance are equally important.

Rather than treating employee AI usage exclusively as a compliance problem, businesses may need to recognise it as evidence of genuine demand for more efficient ways of working.

Understanding why employees turn to unauthorised tools could help organisations identify opportunities to improve their approved technology offerings.

At the same time, clear accountability, appropriate training and transparent data handling policies remain essential.

Ultimately, Acora’s research highlights an uncomfortable possibility for businesses attempting to slow their AI adoption.

The decision to pause an official AI rollout does not necessarily mean employees will stop experimenting with the technology.

Without effective governance and practical alternatives, organisations risk creating an environment where AI usage continues to expand, but increasingly outside their visibility and control.

The challenge for business leaders is therefore not simply deciding whether to adopt AI, but ensuring that adoption happens safely, transparently and with sufficient oversight.

The post UK Searches for Hiding AI Use at Work Surge 489% as Businesses Pause Adoption appeared first on Tech Trends.