Advertisement

Most people don’t trust AI, new KPMG study shows

Most people don’t trust AI, new KPMG study shows

TL;DR

  • A new KPMG study shows that 54% of people are hesitant to trust AI, with trust levels dropping sharply from 63% in 2022 to 46% in 2025.

  • Safety, ethical concerns, and fears about bias, misinformation, and job loss are the main reasons people are wary of AI adoption.

  • AI-powered cybercrime is on the rise, with hackers using tools like ChatGPT and Google Gemini to carry out phishing and malware attacks.

  • Many employees use AI secretly at work without proper training, increasing the risk of errors and weakening overall trust in the technology.

It’s hard to ignore how quickly artificial intelligence is making its way into daily life. From banking apps to hiring software, marketing tools to medical scans, AI is increasingly running in the background of decisions that affect you.

But even as companies continue to adopt it, a growing number of people are getting uneasy about how it’s being used. A new report by KPMG shows that 54% of tech users are hesitant to embrace AI in their workflows.

And it’s not because they don’t see the benefits. They’re concerned about what AI might get wrong, how it’s being governed, and who it might hurt.

According to KPMG, the number of people who see AI as trustworthy has dropped from 63% in 2022 to just 46% today. That’s a sharp decline, and it reflects more than just media headlines or tech fatigue. It points to real, growing concerns about how safe, fair, and secure these systems really are.

The Trust Gap Is Widening

The new global study by KPMG, in partnership with the University of Melbourne, outlines a troubling shift. While many people are excited about what AI can do, fewer are willing to trust it. The drop in trust is not uniform either. It varies depending on how and where AI is being used.

People tend to have more trust in AI when it’s used in healthcare, likely because of the tangible benefits in diagnosis and treatment, as well as a general trust in medical professionals.

On the other hand, applications in areas like HR or generative tools used for content creation draw more skepticism. That’s where trust drops, as people feel the potential for harm or bias is higher, and the benefits are less clear.

Most people don’t trust AI, new KPMG study shows

The hesitation around AI is largely driven by safety and ethical concerns. People aren’t just worried about mistakes or glitches. They’re worried about systems making biased decisions, automating job losses, spreading misinformation, or collecting personal data without proper oversight.

The report shows that 57% of those surveyed either don’t believe current regulations are sufficient or are unsure if existing laws can keep people safe. That’s a big problem, especially when AI is being rolled out at scale across sectors like education, finance, and government.

It’s also worth noting that views on AI vary by region. In emerging economies like Kenya, Nigeria, and India, people tend to feel more optimistic. In these places, AI is often linked with innovation, new job opportunities, and growth.

But in advanced economies like the US, UK, and Germany, there’s more caution. People in these countries report a near even split between excitement and concern.

Cybercrime Is Adding Fuel to the Fire

Trust in AI is also being dragged down by a new kind of threat: AI-powered cybercrime. A report by Check Point Research warns that hackers are now using AI tools like ChatGPT, Google Gemini, and Microsoft Copilot to scale up their attacks.

These tools are helping criminals write more convincing phishing emails, automate malware development, and create synthetic media for scams.

In Kenya, the Communications Authority raised alarms late last year over the growing use of AI in cyberattacks. The agency warned that AI is making it easier for cybercriminals to create fake messages and trick people into giving up sensitive information.

As CA Director-General David Mugonyi put it, attackers are “leveraging AI and machine learning to automate the creation of phishing emails and other types of social engineering.”

A separate study from Ben Gurion University backed this up. Researchers found that most major AI chatbots, including those built by OpenAI and Google, could be tricked into producing dangerous responses. These included instructions for hacking, making weapons, or bypassing security systems.

What Needs to Change

If trust in AI is going to recover, companies, governments, and developers need to do more than talk about ethics—they need to act. First, they have to be transparent about how their systems work, what data is being used, and how decisions are being made.

Users deserve to know when they’re interacting with AI and how it’s influencing outcomes.

There also needs to be clear, enforceable rules about what AI systems can and can’t do. That includes regulation to prevent harmful uses, and accountability for when things go wrong. It’s not enough to say the tech is improving. People need to see that safety, fairness, and accuracy are being taken seriously right now.

Training is another big gap. Most workers are using AI tools without any formal instruction. Companies need to provide that training, not just on how to use the tools, but on when to question them and how to spot errors or bias.

And finally, the tech community needs to build systems that are safe by design. That means addressing bias, securing systems against misuse, and stress-testing models before they go live.

If those changes don’t happen, trust will continue to erode. And with it, the long-term potential of AI may become harder to realize, not because the technology failed, but because the people using it stopped believing it was safe.

Muchiri Avatar

Leave a Reply

Your email address will not be published. Required fields are marked *