Half of UK firms fear staff are feeding company secrets to AI
Half of UK firms fear staff are feeding company secrets to AI, according to a startling new study. Nearly nine-in-ten UK businesses report unauthorised AI use at work, while concerns grow over employees sharing sensitive company data.
Amid the early hype of artificial intelligence, the feverish rush seen in many organisations has led to a number of ‘unforeseen consequences’ in the following years. Research regularly finds that bosses are pushing their staff to use AI in every area of life – though few are actually willing to invest in their own technology. So, while one study of 1,295 business leaders by AI Resume Builder found that 58% of firms required staff to use the tools – and one-in-ten of those said they would fire anyone refusing – another study by the British Chambers of Commerce found that only 10% of UK firms which use AI rely on bespoke or proprietary systems, while the vast majority use off-the-shelf, generic tools.
Exactly what else bosses thought would happen in this case – especially with 86% of UK employees saying they have not been adequately trained for using AI – is a mystery. But now it seems the obvious chickens are coming home to inevitably roost, in a completely avoidable cyber security crisis.
Well, if it isn’t the consequences of my actions
A new study from Turbotic polled more than 1,000 senior decision-makers involved in setting and assessing AI projects in UK businesses, also shed light on the prevalence of ‘shadow AI’, where employees use AI tools without formal approval or oversight. The survey found that generic AI solutions, including simply feeding prompts to the likes of ChatGPT or Claude, had unsurprisingly become pervasive through organisations of all sizes.
A total of 88% of respondents said they believed external AI tools of this kind were being used within their organisation. Put-upon IT departments, pressured into ‘vibe coding’ showed the highest reported use of unauthorised AI at 36%, followed by 30% of customer service respondents, and marketing and sales following at 28% and 27%.
On top of the shoddy service this puts customers in line to suffer from – with hallucinations creeping into every form of contact they have with a company – this also presents a major issue when it comes to privacy. A majority of 51% of businesses now fear employees are feeding sensitive company data into unauthorised AI tools, while they struggle to maintain visibility and control over the rushed deployment of a technology they clearly didn’t understand in the first place.
This lack of comprehension might be illustrated by the fact that – despite its intensive use, and the lack of investment in proprietary AI on many firms’ parts – some 27% of bosses (all of whom have direct involvement in their organisation’s AI initiatives and deployment decisions) said they were not familiar with the term ‘shadow AI’ before the survey. According to the researchers, this shows “that awareness is lagging behind adoption.”
Leaders must start leading
Shutting the stable door after the horse has bolted, many bosses also noted a number of other key worries they might have been better off weighing up a few years ago. These included regulatory or compliance breaches (38%), employees using AI without management oversight (36%) and inaccurate or unreliable AI-generated information influencing business decisions (35%).
The research highlights the challenges that can emerge as businesses move AI from experimentation into wider use, with some 64% saying most or all of their organisation’s AI pilot projects had progressed into day-to-day business operations. The question is, what should firms do now? For Theodore Bergqvist, CEO and co-founder of Turbotic, Shadow AI is showing how quickly adoption can outpace visibility and oversight, and this is where businesses need to pay closer attention.
Bergqvist explained, “What did surprise us is that, as adoption is increasing, just how many organisations don’t know which tools their people are using, or even what the term ‘shadow AI’ means. That’s a huge problem when it comes to properly managing the risks that come with AI use, and encouraging more adoption will only make the problem harder to manage. If organisations want AI to become part of everyday business, they need to first treat visibility, governance and trust as part of deployment, not something to bolt on afterwards. And this responsibility must sit not with employees, but with leadership, who need to be creating an environment where people know what they can use, what they can’t, and why.”
