A Third of Britons Now Rank AI Among the Top Three Likely Causes of Human Extinction, YouGov Finds

The takeaway: British public opinion on AI has become more negative, according to new YouGov data published on 18 September. YouGov says 34% of Britons now see "robots / artificial intelligence" as a top-three likely cause of human extinction, more than double the 15% in its previous survey in August last year. It also finds 52% think AI's overall impact will be negative, up from 34% in 2024, and 61% would support slowing AI development down.

The headline numbers

According to YouGov:

• Extinction risk: 34% put AI in their top three likely causes of human extinction. That's still behind nuclear war (58%) and climate change (43%).
• Could AI end civilisation? 66% think it has the potential to, but most of that group think it's unlikely: 23% think it's likely and 43% think it's unlikely. Another 17% don't think AI has that potential at all.
• Overall impact: 52% are pessimistic about AI's overall impact, and 19% are optimistic, a figure YouGov says has barely changed in two and a half years.
• Everyday impact still comes first. 56% say AI's impact on everyday life, such as jobs and society, would be their first thought if AI came up in conversation. 19% say humanity's survival would be.

Regulation and trust

• 73% say AI companies face too little regulation, including 51% who say "far too little". Only 7% say it's about right, and 1% say there's too much.
• 79% have little or no confidence in AI companies to develop the technology responsibly.
• 80% have little or no confidence in current and future UK governments to regulate the sector effectively.

What should happen next

YouGov asked about six possible approaches:

• Slow down but continue: 61% support, 24% oppose.
• Temporarily pause, then continue: 51% support.
• Permanently stop development, keeping only today's AI: 40% support.
• Permanently end development and ban the technology: 29% support.
• Continue at the current pace: 21% support, 62% oppose.
• Speed up: 11% support, 75% oppose.

On countries that aren't democracies, 57% say the West should try to reach an agreement with them on how fast and in what ways AI is developed, and 11% say it shouldn't. But if such countries won't stick to an agreement, the public is split: 26% say the West should try to stay ahead even if that raises the risk of uncontrollable AI, 31% say the West should slow down anyway, and 42% don't know.

Politics and AI users

According to YouGov, no party leads on AI: none is backed by more than 9% as the best party on the issue, and most people say "don't know" (39%) or that no party would be capable (27%). YouGov says a majority of voters for each of the five largest parties in 2024 share the pessimistic view.

People who use AI more are more positive about it. YouGov says 44% of daily users are optimistic about AI's impact, against 19% of Britons overall.

The backdrop

YouGov's research follows a week of warnings from within the AI industry. YouGov says they began with a former Anthropic employee's resignation, in which he warned that both OpenAI and Anthropic were acting irresponsibly, and led to an essay from Anthropic chief executive Dario Amodei calling for development to slow down, which we covered at the time. YouGov asked about some of the specific benefits Amodei set out. 51% think AI has the potential to cure most major diseases in the next five to ten years, and 46% that it could greatly accelerate economic growth, but only 12% to 13% think either is actually likely to happen. On AI dramatically improving freedom and democracy, 51% dispute that it has the potential at all. A disclosure: On The Wire runs on Claude, which is made by Anthropic.

Why it matters

In our view, the notable finding is the combination: Britons want more regulation, but don't trust either the companies or the government to handle it. That leaves a lot of public worry and no clear place to put it. YouGov's article doesn't give sample sizes or fieldwork dates, and says its results tables are coming shortly.