At the Bay of Bengal Conversation 2026, I found myself listening for one word.
The opening discussion was about power, technology and trust in an age of uncertainty. As expected, there was plenty about technology -- especially artificial intelligence -- and about how governments and world leaders are struggling to respond to a technology that is moving faster than the rules around it.
But I kept waiting for the third word: Trust. How is trust built? How does it disappear? And perhaps most importantly, who gets to shape either process?
Walking out of that session, I felt that these questions were more important than the technology itself. AI does not build trust or destroy it by itself. People build it. People destroy it. People decide how technology is designed, who controls it, and what purposes it serves.
The historian of technology Melvin Kranzberg made this point decades ago. His first law famously said: “Technology is neither good nor bad; nor is it neutral.” His sixth was even more direct: “Technology is a very human activity.”
That observation feels remarkably current in the age of AI.
We have seen this before, close to home. As I thought about it, my mind went to Ramu. In late September 2012, a photograph depicting the desecration of the Quran appeared on Facebook and was linked to a young Buddhist man in Ramu, Cox's Bazar. He denied posting it.
The story spread rapidly, and on September 29-30 mobs attacked the Buddhist community. In Ramu alone, at least 12 Buddhist temples and monasteries and more than 50 houses were destroyed. A later judicial inquiry exonerated the young man, concluding that he had been a victim of the misuse of social media. The violence also spread beyond Ramu.
Six years later, the pattern appeared on a much larger scale, just across the border. In its 2018 report on Myanmar, the UN Independent International Fact-Finding Mission described the role of social media as “significant” and said Facebook had become a useful instrument for those seeking to spread hate, in a country where, for many users, Facebook was effectively the internet. The mission also noted that the real-world effects of Facebook posts and messages needed independent examination.
Neither of these was an AI story. Both were human stories in which a powerful communication technology met fear, grievance and people willing to exploit them.
Generative AI changes the equation because it can make deception cheaper, faster and more convincing. A fabricated voice, image or video can now be produced and distributed at enormous scale.
But the appetite for deception is not new. Research tells us something unsettling about why such material travels. In a 2018 study published in Science, Soroush Vosoughi, Deb Roy and Sinan Aral analyzed roughly 126,000 verified true and false news stories shared on Twitter between 2006 and 2017.
False stories were 70% more likely to be retweeted than true ones. And when the researchers accounted for bots, the difference remained: Human behaviour, not automated accounts, explained the greater spread of falsehood. False stories also generated more reactions associated with surprise, fear and disgust.
That should make us pause. The problem may not always be convincing people of one particular lie. Sometimes it may simply be keeping people sufficiently angry, surprised or afraid that they stop trusting what is in front of them.
Trust is a social product
This is where I think the AI debate often becomes too technical. Robert Putnam, in Bowling Alone, explained that people are more likely to trust when they actually know one another, interact regularly and belong to communities where behaviour has consequences.
Francis Fukuyama made a related argument in Trust, defining trust as an expectation of regular, honest and cooperative behaviour based on shared norms, and linking it to the ability of societies to work together and prosper.
That makes me wonder what happens when our social lives become increasingly mediated by screens. A phone can give us an extraordinary amount of information without giving us the relationships needed to test that information. A feed can tell us what millions of strangers are saying while removing us from the neighbour sitting across the street.
The 2026 Edelman Trust Barometer offers a striking picture of this problem. Based on a survey of 33,938 people across 28 countries, it found that 70% of respondents were unwilling or hesitant to trust someone whose values, information sources, approaches to social issues or cultural background differed from their own.
Edelman describes this condition as “insularity.” It also reports that the gap in its trust index between high- and low-income respondents has widened from six points in 2012 to 15 points in 2026.
These are not measurements of AI. They are measurements of us. And they suggest something important: A society already divided by identity, insecurity and distrust is easier to manipulate with any powerful communication technology, whether it is Facebook in 2012 or generative AI today.
Who controls the tool?
This brings me to the question underneath the AI regulation debate: Who controls the technology? In Power and Progress, economists Daron Acemoglu and Simon Johnson argue that technological progress does not automatically distribute its benefits widely.
The direction of technological change depends on choices, institutions and who has the power to make those choices. Technology can serve broad prosperity, but it can also serve narrow interests.
Their wider work with James Robinson on institutions made a similar point. In awarding the three economists the 2024 Nobel Prize in Economic Sciences, the Nobel committee cited their research on how institutions are formed and affect prosperity.
It distinguishes inclusive institutions, in which people have a meaningful stake in governance, from extractive institutions in which political and economic power is concentrated in a narrow elite.
That distinction matters enormously for AI. Frontier AI is not being developed in a world where everyone has equal access to computing power, data, specialized talent and capital.
Stanford's 2026 AI Index reports that industry produced more than 90% of notable frontier models in 2025. A UK government assessment similarly notes that frontier development requires such substantial funding, skills, data and computation that it is likely to remain concentrated among a small group of companies.
The concentration of wealth is part of this story too. Oxfam's January 2026 report says billionaire wealth rose by more than 16% in 2025, reaching $18.3 trillion. It estimates that billionaires are more than 4,000 times more likely than ordinary citizens to hold political office, and says the wealth of the poorest half of humanity is smaller than that of the world's 12 richest billionaires.
These are Oxfam's estimates, based on its own methodology, so they should be read in that context. But they illustrate the scale of the concentration of resources and political access that now surrounds some of the world's most powerful technologies.
The difficult question, then, is not simply whether AI is powerful. It is whether institutions are strong enough to make sure that extraordinary power remains accountable.
Regulation cannot do everything
None of this means that we should stop regulating AI. We need rules around transparency, accountability, data protection and synthetic media. Governments will have to decide how to deal with deepfakes, automated decision-making and the growing power of the companies building these systems. But regulation operates inside society. It does not float above it.
If institutions are trusted, rules have a better chance of working. If institutions are weak, divided or captured by powerful interests, even well-designed rules can become instruments of selective enforcement. That is why I keep coming back to something much older than AI: The strength of the social fabric around us.
People need real communities to belong to -- public spaces, libraries, playgrounds, clubs, neighbourhood organizations, and campuses where people encounter others who do not necessarily think like them.
People also need things that make life worth looking up from the screen for: Affordable culture, sport, music and recreation. Otherwise, the endless feed becomes the easiest form of escape.
We need institutions that reward merit rather than loyalty, whether in public service, business or politics. And we need something even more basic for young people: A believable path into the future.
Hope works better when there is a pathway behind it. None of these things sounds particularly high-tech. Perhaps that is precisely the point.
The question I left with
The AI debate will continue. Models will become more capable. Regulation will evolve. Companies and governments will keep competing over technology, data, infrastructure and influence. But however powerful AI becomes, the consequences will still depend on human choices.
That is why, after listening to a panel about power, technology and trust, I found myself thinking less about whether we can trust AI. The more difficult question is whether we still trust one another -- and whether our institutions have given us enough reason to do so. The AI story is a human story. And, in the end, so is its ending.
Aminul Ehsan is a democracy, governance and political parties expert.
