AI is now ubiquitous in UK startups, and so are concerns about its use. Half of founders have shared sensitive information with a public AI tool in the past month. majority A new study in collaboration with Sage has raised concerns about the privacy of these models.
Startups looking to hedge public AI risks with custom stacks also face roadblocks, with 62% wasting AI budgets repairing broken connections between orphaned tools. This leaves founders with a dilemma: accept the risk or absorb the cost.
To help navigate this challenging landscape, we proudly partnered with Sage to release the definitive 2026 report and handbook for founders: The Startup AI Paradox.
Based on a survey of more than 400 founders, managers and executives of UK startups, this report reveals the challenges perceived by today’s executives using AI and explores potential solutions. This is what we revealed.
90% of small businesses using AI are concerned about their private data being used to train global models
Startups are increasingly relying on public AI tools to quickly increase productivity. 1 in 2 founders admit to using chatbots to leak confidential company information Over a quarter of them are becoming a weekly habit, like ChatGPT and Gemini in the past 30 days.
But this is not due to simplicity. 90% of AI users are concerned that the personal information they enter into tools could potentially train someone else’s modelsand more than a quarter (27%) say they have little or no trust in AI to handle core business processes.
The risks are obvious. Once sensitive corporate information is entered into a public model, it cannot be undone, deleted, or contained. This could result in client data reappearing in another user’s output and subject to GDPR penalties.
This gap is driven by competitive pressures. As business environments rush to incorporate AI into daily workflows, standing still can feel like a bigger risk than ever. Our research also found that 83% of business leaders feel overwhelmed by the number of choices available to themAs a result, many of them end up defaulting to what’s available.
Whatever the information supporting their decision, the findings reveal the fact that founders are aware of the risks of using public AI and continue to use it anyway.
Companies face a catch-22 when it comes to choosing how to use AI.
For founders who are attracted to the speed and accessibility of public AI, using it securely is not impossible, but just cumbersome. Publishing tools have no data security guarantees, and founders must anonymize sensitive people, wipe internal documents, and remove sensitive details before a single prompt is sent.
Our findings highlight how tedious founders find this process. Half of respondents (56%) spend a moderate amount of time preparing data manuallyand almost a quarter (23%) feel completely self-defeating.
To get around these issues, some companies are building their own AI technology stacks, where each model can handle different information streams, from customer data to financials, giving founders more control over where sensitive data goes.
However, while a step in the right direction, enterprise-built AI infrastructure can increase costs while improving efficiency. 62% of companies have dedicated AI budget reporting Waste a significant amount of time trying to repair broken connections Among their tools.
At the end of the day, trade-offs remain whether companies absorb the risks of public AI tools or invest in fragmented custom stacks. Neither method solves the problem, but our findings point to another option.
See the big picture with expert-driven AI playbooks
No matter where your company is in its AI journey, help is readily available. Our free report “The Startup AI Paradox” A comprehensive AI handbook for founders to navigate the trade-off between speed and security.
Featuring insights from Startups 100 winners, a starter checklist for founders looking to develop an AI policy, and legal guidelines explained in clear English, all designed to take the guesswork out of adopting and thriving with AI.
We will also discuss solutions that can permanently resolve this conflict. Download “The Startup AI Paradox” for free and get the full results from the study in collaboration with Sage.
