Rishi Sunak: AI's Future Depends on Adoption and Public Trust

The former UK prime minister calls for independent checks inside AI labs and external testing before frontier models are released.

RSRishi SunakSpeaker · Prime Minister
Channel
Bloomberg Live
Published
Nov 18, 2026
Video
12:00
Read
5 min
Original
English
Open on YouTube
Scroll to read and the video shrinks to the bottom-right corner. Captions follow your reading language.
Reading languageAvailable in 2 languages
You're reading the original English.
SUMMARY

In 30 seconds

Rishi Sunak argues that the countries and companies that spread AI through everyday life may benefit more than those that merely invent it. But he says public trust could limit deployment in democracies unless people see that the technology will help rather than harm them. Drawing on his experience convening the 2023 Bletchley Park AI safety summit, he proposes independent evaluators inside labs and external tests of frontier models before release, ideally with government oversight. He also points to India's digital infrastructure and a low-cost personalized learning program as examples of adoption's potential.

  1. 1The race to spread AI may matter more than the race to invent itSunak argues that countries gain from how effectively they deploy a technology. He points to the Dutch Republic's printing industry as an example of adoption yielding economic value elsewhere than the invention's birthplace.
  2. 2Sunak sees India as well placed to adopt AIHe cites its AI-literate workforce, digital public infrastructure and startup ecosystem. He describes the Mindspark tutoring program as an example of how inexpensive personalized education might reach children in India and Africa.
  3. 3Public consent could become AI's limiting factorSunak contrasts optimism in India with negative sentiment in the UK and US. He argues that people need evidence AI will empower them rather than replace or harm them.
  4. 4Waiting for a serious accident could bring a sweeping backlashSunak warns that the scale of a future incident is unknown and that a disaster might prompt regulation that halts development and deployment. He argues for acting before that happens.
  5. 5Independent evaluators should see how labs develop their modelsHe proposes independent reviewers inside labs and external testing of every frontier model before release. Government involvement is his preference because of democratic oversight and national-security knowledge.

Takeaways

  1. Sunak argues that widespread adoption, not only being first to invent, determines who benefits from a technology.
  2. He says public consent could limit AI deployment if people fear replacement or harm.
  3. He proposes independent evaluators inside labs and external testing before each frontier model is released.
  4. Sunak says targeted rules for severe risks should preserve competition and growth.
FULL TEXT

Read the edited text by chapter

Transcript only
01

Why AI may spread faster than past technologies

Sunak compares AI with the Industrial Revolution, the Enlightenment and the printing press. He argues that its combination of technology, discovery and access to knowledge could accelerate change.

Good morning, and thank you for being here on the first morning session of the final day. Your energy has been inspiring throughout the forum. I'm delighted the Bloomberg New Economy Forum is in Delhi; my daughters are also pleased that I'll be returning with pre-Diwali presents.

Demis Hassabis has said AI will have ten times the impact of the Industrial Revolution in half the time. A colleague at Google has compared it to the Industrial Revolution and the Enlightenment combined: a new technology alongside a new way of thinking and discovering.

I would add the printing press, which democratized knowledge and information. AI combines technology, new methods of discovery and the spread of knowledge. I believe that combination will make the coming pace of change faster than anything we have experienced.

Consider the time it took technologies to reach 100 million people. The telephone took 70 years, electricity 50, the personal computer 15 and the internet seven. ChatGPT took just two months. That is the speed at which we are now operating.

02

The advantage of adopting a technology

The Dutch Republic's printing industry illustrates Sunak's argument that invention alone does not determine who gains. He describes India as well positioned to spread AI because of its workforce, infrastructure and startups.

Headlines focus on the race to be the first company or country to achieve artificial general intelligence. I think the more important race is for everyday AI: spreading the technology's benefits throughout the economy and society. Leadership depends not only on invention but also on how effectively a country deploys and adopts it.

The printing press is an example. Johannes Gutenberg invented it in Mainz, but Professor Geoffrey Ding has pointed out that the Dutch Republic extracted much of its economic value, becoming a publishing powerhouse in the 17th century, while Gutenberg went bankrupt.

The Dutch had less powerful guilds than the German states, so more people could set up printing businesses. They also had fewer restrictive censorship laws, and Amsterdam's financial markets helped people manage paper and printing costs. The lesson I draw is that countries, companies and organizations must adopt the technology.

India is well placed: it has an AI-literate workforce, digital public infrastructure such as UPI, a vibrant startup ecosystem and a culture of frugal innovation. It sent Chandrayaan to the moon for less than the cost of making Interstellar. India recently moved ahead of the UK into third place in Stanford's ranking of global AI powers.

The AI summit hosted by Prime Minister Modi in February highlighted how the technology might increase growth and raise living standards in the developing world. It is already providing solutions in agriculture, healthcare and education at a fraction of what they would have cost a generation ago.

03

Personalized education and the case for public consent

Sunak cites Mindspark as an example of affordable personalized learning. He says trust and consent are necessary for AI deployment in democratic societies.

Take Mindspark, developed in India and now teaching half a million pupils in India and Africa. It provides personalized lessons for a few dollars a month; I understand the children's rate of learning has doubled. It can use a simple tablet with preloaded content rather than fast broadband or an expensive laptop, drawing on 20 years of research and five billion student interactions.

The AI conversation in my country, and even more in the US, is much more negative than in India. That concern must be understood. The most important condition for developing AI in a free society is public consent: without it, having all the computing power in the world will not be enough.

Trust threatens to limit AI more than technology until we can show people that it will empower rather than replace or harm them. Public opinion can shift quickly; a halt to AI development can move from a fringe position into the mainstream. Look at how the political mood has turned against data centers in the US.

04

Safety checks before an accident

Sunak argues that waiting for harm could trigger a broad backlash. He proposes independent oversight within labs, external model testing and predictable rules targeted at severe risks.

I have always thought AI progress and safety go hand in hand. That is why I made safety the focus of the 2023 world leaders' summit at Bletchley Park. One useful piece of advice I received as a minister was to ask what you would do after something goes wrong, and do it now.

There is a creeping fatalism that agreement on safety will come only after an incident with real-world consequences. I think that is risky and morally flawed: we do not know how bad an accident could be. A disaster could also lead to sweeping regulation that halts AI development and deployment. We should get ahead of it, with industry moving quickly and inviting government in.

First, independent evaluators should operate inside labs to check whether they are following their own safety procedures and to see how models are developed and tested. Ideally, they would come from government for democratic oversight, but independence and access to the labs are the essential features.

Second, every frontier model should be tested externally before release. I would again prefer governments to carry this out, both for democratic accountability and because they have a broad view of national-security implications.

We also need transparency and clear, predictable rules that customers, investors, companies and the public can understand and trust. Government involvement can stifle innovation or protect incumbents, so rules should focus on the most severe risks while supporting competition and growth.

05

Choosing how AI improves lives

Sunak describes AI as a tool that can help create new tools. He calls for using it to improve lives as well as productivity.

AI is a tool for discovering and making new tools—an innovation in the method of innovation itself. There are many possible futures. The choice is how to use the technology to raise living standards and empower people, as well as make economies more productive.

More from this channel

15:57EN
Bloomberg Live· Nov 18, 2026Gina Raimondo: AI Safety Cannot Be Left to Countries Acting AloneGina RaimondoThe former US commerce secretary calls for shared AI safety standards and worker retraining as the technology outpaces governments.AI SafetySafety & governanceFuture of Work+2
18:18EN
Bloomberg Live· Nov 16, 2026Druley and Tolani on Diversification as Investors Look Toward 2027Farah Elbahrawy, David Druley, Ekta Tolani · BloombergThe investors weigh concentrated US markets against cash yields and opportunities in venture capital, Europe and the Middle East.InvestingVenture CapitalGeopolitics+1
28:17EN
Bloomberg Live· Nov 12, 2026COP31’s electrification push meets questions about clean power and tradeRebecca Penty, John Ainger, Lili Pike, Selcan Hacaoğlu · BloombergBloomberg reporters examine summit negotiations, China’s industrial plans and Europe’s energy-security dilemma.EnergyGeopoliticsEconomics
Slides14:12EN
Bloomberg Live· Nov 12, 2026Clean energy needs better grids, contracts and supply chainsAlbert Cheung · BloombergNEFBloombergNEF’s Albert Cheung says cheap renewables can grow rapidly, but grids, revenue models and copper supply must keep pace.EnergyEconomicsGeopolitics
20:33EN
Bloomberg Live· Nov 12, 2026Family investors weigh AI disruption against the businesses funding itAbeer Abu Omar, Mishal Kanoo, Shravin Bharti Mittal · BloombergMishal Kanoo warns that family blame can discourage risk; Shravin Bharti Mittal tests new technology against its route to customers.InvestingVenture CapitalCompany Building+1
25:04EN
Bloomberg Live· Nov 12, 2026PIMCO and Blackstone see Gulf financing needs shift toward private capitalAbeer Abu Omar, Farhad Karim, Christian Stracke · BloombergChristian Stracke expects regional demand for private lending; Farhad Karim sees domestic investment as an opportunity rather than a retreat.InvestingGeopoliticsEconomics+1