Public trust in artificial intelligence has eroded sharply since 2019, even as the tools reached billions of people through products like ChatGPT. The 2023 Edelman Trust Barometer reported that faith in AI firms fell to 50% globally, down from 61% in 2019. A separate KPMG survey that year found 61% of respondents were wary about trusting AI systems. These figures define the operating environment for every business deploying AI: skeptical users are harder to convert, and regulators facing skeptical electorates move faster than executives expect.
The European Union's AI Act, formally adopted by the EU Council in May 2024, is the most concrete regulatory answer to this sentiment shift. But the law arrived years after public unease began rising. Pew Research Center reported in 2023 that 52% of Americans were more concerned than excited about AI, up from 37% in 2021. The question now is whether regulation can close the distance between what the public fears and what the tools can actually do.
What the Major Surveys Show
The Ipsos Global Advisor survey in 2022 found that 78% of people across 28 countries were worried about AI being used for deceptive purposes. That figure is remarkably high and consistent across markets. It suggests the dominant public emotion toward AI is not fear of job loss or existential risk, but a more immediate anxiety: that the systems will be used to mislead them.
Edelman's data shows a steep decline in trust from 2019 to 2023, a period that predates the generative AI boom. The figure dropped 11 percentage points before ChatGPT even existed. That slide tracks with high-profile failures, data breaches, and the general sense that organizations were deploying AI faster than they could secure it. KPMG's 2023 finding of 61% wary respondents reinforces the message: by the time generative AI reached mass adoption, the reservoir of goodwill was already draining.
Generative AI Changed the Conversation
ChatGPT reached 100 million monthly active users within two months of its November 2022 launch. That adoption rate was unprecedented, and it put AI in the hands of people who had never used it before. The immediate effect was a surge in both excitement and anxiety.
From abstract to tangible concern
Before ChatGPT, public discussion of AI was largely abstract: self-driving cars, medical diagnosis, chess computers. After ChatGPT, everyone could see a machine write a poem, answer a question, or produce a plausible fake. The Ipsos finding that 78% of people worried about deceptive AI use suddenly had a concrete example. Pew's shift from 37% to 52% concerned Americans from 2021 to 2023 captures the same phenomenon in survey data.
How Attitudes Differ by Country and Demographics
Survey data shows that AI optimism is not evenly distributed. Markets with strong state-led AI strategies, such as China, tend to show higher trust levels. Markets with high media scrutiny of tech firms, such as the United States and many European Union member states, show lower trust.
The demographic pattern is also consistent. Younger people and those with higher education levels tend to be more optimistic about AI. Older people and those with less direct exposure are more likely to be wary. Income also correlates: higher earners see AI as a tool, lower earners see it as a threat to their job. None of these patterns are unique to AI. They mirror attitudes toward earlier automation, which suggests the underlying driver is not the technology itself but the social position of the person evaluating it.
The Applications That Worry People Most
Not all AI applications generate the same level of worry. The Ipsos survey's 78% figure on deceptive use points to a specific fear: AI used to impersonate, mislead, or manipulate. That includes deepfakes, disinformation, and fraudulent customer service calls.
Healthcare and scientific research applications tend to generate more optimism. People are broadly comfortable with AI that helps diagnose disease or accelerate drug discovery. The difference is control: people feel they are the customer of medical AI but the target of deceptive AI. Job displacement sits in the middle. It is a real worry, but it ranks below the fear of being actively deceived. The distinction matters for product design. An AI that helps a doctor read an X-ray faces less public resistance than an AI that writes a campaign ad.
Regulation as a Response to Public Sentiment
The European Union's AI Act, adopted in May 2024, is the world's first comprehensive AI regulation. It classifies AI applications by risk level and imposes stricter rules on high-risk systems. The timing is not coincidental. European regulators had been drafting the law since 2021, but the surge in public unease after ChatGPT's launch accelerated the will to finalize it.
Other jurisdictions are moving more slowly. The United States has issued executive orders and voluntary commitments but lacks a federal AI law. China has introduced rules for algorithmic recommendations and deep synthesis but has not matched the EU's scope. The gap in regulatory pace reflects different governing systems, but also different levels of public pressure. In countries where trust in AI builders is lower, the cost of inaction is higher.
How Companies Are Responding
Technology firms are responding to public worry with a mix of transparency measures and lobbying. OpenAI, the organization behind ChatGPT, has published system cards and usage policies. Microsoft and Google have made voluntary commitments to responsible AI development. But these efforts have not reversed the trust decline. Edelman's data shows faith in AI builders dropping to 50% in 2023, meaning half the global population does not trust the enterprises building the most transformative tool of the decade.
The gap between public understanding and actual AI capability remains the core problem. Many people attribute more agency and intelligence to AI systems than they possess. Businesses that invest in explaining what their AI can and cannot do are likely to fare better than those that lean into hype. But the survey data suggests a simple truth: the public is not worried about the technology. It is worried about the people running it. Until that changes, no amount of regulatory text will rebuild trust.
Key Facts
- Ipsos 2022 concern about deceptive AI: 78% across 28 countries
- Pew 2023 Americans more concerned than excited: 52%, up from 37% in 2021
- Edelman Trust Barometer 2023 trust in AI companies: 50%, down from 61% in 2019
- KPMG 2023 wary about trusting AI systems: 61%
- EU AI Act adoption date: May 2024
- ChatGPT time to 100 million monthly active users: 2 months after November 2022 launch
Public Trust in AI Companies Over Time
| Year | Trust Level (Edelman) | Concern Level (Pew, US only) |
|---|---|---|
| 2019 | 61% | Not measured |
| 2021 | Not measured | 37% more concerned than excited |
| 2023 | 50% | 52% more concerned than excited |
Frequently Asked Questions
Has the EU AI Act changed public trust in AI?
The EU AI Act was adopted in May 2024. As of that date, no major global survey had measured its effect on public trust. The impact remains to be seen.
Which country has the highest trust in AI?
The brief does not provide country-level trust comparisons. However, markets with strong state-led AI strategies, such as China, tend to show higher trust levels in available survey data.
Is public concern about AI increasing or decreasing?
Pew data shows concern increased in the United States between 2021 and 2023. Edelman data shows trust in AI companies declined globally between 2019 and 2023. The trend is toward greater concern, not less.
