As governments and industries worldwide continue to invest in advancing AI technologies, a new survey from KPMG has found that the global public is less trusting of artificial intelligence and keenly aware of the risks associated with its rollout.

The professional services firm’s Trust, Attitudes and Use of Artificial Intelligence report found that although 66% of people are already using AI regularly, and 83% believe AI will result in a range of benefits, less than half are willing to trust it (46%).

Polling 48,000 people across forty-seven countries, KPMG discovered that, compared to its last study conducted just before the release of ChatGPT in 2022, people have become less trusting and more worried about AI adoption.

In the UK, just 42% of people said that they trusted AI systems, with more than half (58%) worried about their use, driven by a high fear of jobs being lost to emerging tech (86%).

Around the world, four in five people are now concerned about the risks of AI, with two in five reporting personally experiencing negative impacts due to the technology.

The view that the benefits of AI outweigh the risks has decreased from 50% in 2022 to 41% today, while concerns about those risks have rocketed, most notably in the Netherlands, up from 67% being concerned in 2022 to 85% in 2024, and Germany, rising from 65% to 79%. 

Drilling down into these perceived risks, cybersecurity surfaced as a top concern across almost all countries, with 85% of global respondents worried about AI-driven hacking and malware. 

A similarly high number (80%) flagged AI-powered misinformation and disinformation as a major threat, with 64% fearing that elections are being influenced by AI-generated content and bot activity, while 80% were most worried about job loss due to AI – something personally experienced by 42% of respondents.

Despite their anxiety, around three in five (58%) of employees are intentionally using AI, a third (31%) of them using it weekly or daily, due to the range of benefits AI is reportedly bringing to their work, with almost half (48%) reporting that AI tools have increased revenue-generating activity.

KPMG found employees are becoming more reliant on these tools, with around half of those surveyed regularly choosing to use AI to complete work, rather than collaborating with peers or supervisors, despite 82% having fears around deskilling.

Employees admit to using AI in ways that contravene their company policies, including uploading sensitive company information into free public AI tools like ChatGPT, leading 57% to hide their use of AI and present AI-generated work as their own.

Sixty-six percent of employees even said that they cannot complete their work without AI assistance.

The survey paints something of a complex picture, with AI being heavily relied upon but equally mistrusted, something employees feel pressure to use as a necessary evil in keeping up with peers, even if it makes them uncomfortable to use.

Recommended reading

• Can AI Ever Replace Human Musicians?
• The Biggest LLMs Are Generating Vulnerable Code by Default
• Are Enterprises Sitting on Untapped Tech Talent?

However, the negative impacts of AI could be mitigated through better training, regulation and governance. Only 47% of employees say they have received AI training, despite 60% of organisations using AI providing it, while just 40% say their workplace has a policy or guidance on generative AI use. 

More broadly, there is a clear public demand for more international law and regulation and for industry to partner with government to mitigate AI risks. Only 43% of the global public believe existing laws and regulations around AI are adequate, with 70% agreeing that more regulation is required before the technology completely overwhelms society. 

“Organisations have a clear role to play when it comes to ensuring that AI is both trustworthy and trusted,” said David Rowlands, global head of AI at KPMG International.

“People want assurance over the AI systems they use which means AI’s potential can only be fully realised if people trust the systems making decisions or assisting in them.”