Majority of Americans see risk AI could destroy humanity: poll
AI illustration. Photo: AA
A majority of Americans believe advanced artificial intelligence poses at least a moderate risk of eventually destroying humanity, according to a Politico poll released on Wednesday, highlighting growing public concern over the rapid development of AI technology.
About two-thirds of respondents said there is at least a moderate risk that advanced AI could destroy humanity, including 17% who said it is “almost certain” and 20% who described the risk as “significant". Another 26% characterised the risk as moderate.
The survey found concern across political, gender and age groups. Among voters in the 2024 presidential election, 70% of those who voted for Democratic candidate Kamala Harris saw at least a moderate risk, compared with 60% of Donald Trump voters. President Trump has downplayed the fears about AI.
Americans also expressed greater support for slowing AI development than for continuing to build increasingly advanced systems.
Read: OpenAI's rogue agents probed Hugging Face for weaknesses two months before major hack
According to the poll, 48% said AI is “good enough”, and development should be paused to reduce risks, while 31% favoured continued development because advanced AI could bring significant benefits. Among Harris voters, 58% backed a pause, compared with 44% of Trump voters.
The findings come as prominent AI industry figures have publicly warned about the technology's potential dangers. Anthropic CEO Dario Amodei and OpenAI CEO Sam Altman, alongside billionaire Elon Musk, have called for slowing the development of advanced AI.
About half of Americans surveyed said AI leaders warning that the technology could ultimately kill humanity genuinely believe their warnings. By contrast, 26% said such statements are intended as marketing to make AI products appear more powerful or important.
Trump has continued to emphasise AI development as an economic and technological priority, while dismissing warnings about existential risks. Republicans and Democrats in Washington are nevertheless debating how to balance AI safeguards with the need for the US to remain competitive with China.
The poll was conducted online by London-based Public First on September 13-15 among 2,064 US adults. It has a margin of error of plus or minus 2.2 percentage points, with higher margins for smaller subgroups.