How might AI technology serve as catalyst for global peace?

An extraordinary occurrence transpired within the chambers of the United Nations Security Council. Sam Altman, Chief Executive Officer of OpenAI, and Dario Amodei, Chief Executive Officer of Anthropic—executives representing two organizations that maintain intense competitive rivalry within the artificial intelligence sector—articulated their perspectives utilizing remarkably congruent terminology.

Speaking at the UN Security Council on September 23, Amodei warned that AI could pose a risk to humanity if it is not properly governed.

Altman, meanwhile, cautioned that control over the future could eventually shift from humans to AI. In other words, the very people building this technology were standing before the world’s most powerful security forum and effectively saying: regulate us.

On the subsequent day, United States President Donald Trump and Chinese President Xi Jinping convened at the White House. Prior to their meeting, Trump characterized artificial intelligence as super intelligence and designated it as the primary matter for discussion. Through a statement posted on Truth Social the preceding day, Trump explicitly communicated his opposition to implementing limitations on artificial intelligence technology.

He also said he did not believe AI posed a threat to human civilisation.

Yet after the meeting, a very different message came from Chinese President Xi Jinping. He said China would work with the United States to improve AI governance. Xi stressed that both countries had the capacity and responsibility to develop and manage AI, and that they must ensure AI remains under human control.

The much-discussed meeting, however, produced neither a binding agreement nor a joint declaration. AI developers are calling for guardrails, while those who possess the political authority to impose them are, for now, speaking mainly of goodwill and cooperation. This raises a fundamental question: can AI genuinely contribute to world peace, or is it becoming the most dangerous weapon of a new era?

After nearly three decades of observing the telecommunications and information technology sectors, one conclusion has repeatedly struck me: technology itself does not create either peace or war. The mobile phone has enabled farmers in remote villages to check market prices. The same mobile phone, however, has also been used to spread rumours over trivial incidents, sometimes leading to attacks on homes and businesses. The same principle applies to AI, but on a far greater scale. AI is not merely a technology that carries messages or responds to human instructions. It can increasingly analyse information and make decisions with limited human intervention. And it is humans who have endowed technology with this increasingly decision-making capability.

It is therefore important to make one thing clear from the outset: the fundamental causes of war are political, economic and human. No algorithm can eliminate the desire for power, territorial disputes or ethnic hatred. But when used responsibly, AI can make the process of building peace faster, more evidence-based and more inclusive. Those possibilities cannot simply be ignored.

Where AI can work for peace
One. Early warning of conflict. Preventing a fire is always preferable to putting one out after it has started. The VIEWS project, a joint initiative of Sweden’s Uppsala University and Norway’s Peace Research Institute Oslo, analyses economic indicators, food prices, climate data and social media information to regularly assess where the risk of political violence is increasing. The critical question is whether there is sufficient political will to act on such warnings. If alerts simply remain buried in files, even the world’s most sophisticated algorithm is of little use.

Two. Bringing ordinary people’s voices into peace negotiations. Peace negotiations are usually conducted by those holding the guns, while those who suffer from the guns are left outside the room. AI-assisted systems, however, can rapidly collect, organise and amplify the voices of people living amid conflict, potentially putting pressure on those holding the trigger to reconsider their actions. The United Nations, for example, has used AI-assisted digital dialogue initiatives in Yemen, Libya and Iraq to collect and analyse the views of thousands of citizens. Such initiatives can help identify areas of consensus as well as points of division.

Social media once provided an important platform for women, young people and marginalised communities to make their voices heard. But social media algorithms have largely failed to prevent the spread of rumours, defamatory content and misinformation.

Platforms such as Facebook and YouTube have not been particularly effective at distinguishing between truthful and false content. On the contrary, the weaknesses of their algorithms have often allowed defamatory, fabricated and, quite simply, harmful content to go viral. Important and constructive content does not necessarily receive the same level of reach. As a result, social media has increasingly struggled to serve as a reliable source for public opinion research or serious information analysis. The irresponsible operation of such platforms has also placed countless innocent people at risk in different parts of the world.

AI, however, has the potential to assess information in a more contextual manner. An AI agent can analyse the nature and purpose of a user’s instructions and, in certain circumstances, identify potentially harmful requests and advise the user against them. It can also gather data from a large number of reliable sources within a very short period and produce a structured analysis.

Such analysis could help everyone from ordinary users to policymakers make better-informed decisions.

That does not mean AI is incapable of making mistakes. It certainly is. But while social media platforms continue to struggle with algorithmic weaknesses and misinformation, AI systems are rapidly improving their analytical capabilities. Anthropic’s latest report, for example, documented how and where AI agents had been used for harmful activities while also demonstrating the company’s efforts to identify and address such misuse. This kind of institutional responsibility has not always been evident in the operations of major social media platforms.

There is also an emerging discussion about whether AI can help make the algorithms of social media platforms more responsible. If social media companies continue to treat harmful content as a commercially valuable product, economic incentives may ultimately push technology developers to train AI systems to operate in more responsible and constructive ways. This is where I see significant potential for AI to become an important technology in the effort to preserve global peace.

1. Breaking down language barriers. AI-powered real-time translation is making direct communication easier between communities speaking different languages and international peacekeepers.

For Bangladesh, one of the leading troop-contributing countries to UN peacekeeping missions, this is more than a technological development. When a Bangladeshi peacekeeper in a remote African village can understand what local residents are saying, the risk of deadly misunderstandings can be reduced.

1. Monitoring ceasefires and agreements. AI can analyse satellite imagery to identify troop movements, the transportation of weapons or the destruction of villages. When violations of an agreement can be independently documented, mutual confidence becomes easier to build. Without trust, a peace agreement remains little more than a document.

2. Addressing the roots of conflict. Poverty, food insecurity and competition over water and land have contributed to many conflicts.

The World Food Programme’s HungerMap Live uses AI and near-real-time data to monitor food insecurity. More accurate forecasting of floods, droughts and cyclones could become an important tool for climate-vulnerable countries such as Bangladesh by helping prevent displacement and competition over scarce resources.

The risks must also be taken seriously
Alongside AI’s potential to contribute to peace, its risks must be examined with equal clarity. UN Secretary-General António Guterres has warned this week about autonomous lethal weapons—often referred to as “killer robots”—that operate without meaningful human control. In several recent conflicts, AI-assisted targeting systems have raised serious questions about civilian casualties and accountability.

Ukrainian President Volodymyr Zelenskyy, whose country has been at war for more than four years, told the General Assembly that if human control over machines is not established quickly, it could soon be too late.

The possibility of losing control over AI agents is no longer confined to science fiction. Australian Prime Minister Anthony Albanese has said that in June, an OpenAI agent accessed confidential files on the country’s Medicare statistics portal despite repeated attempts to block it. OpenAI reportedly informed the Australian government 84 days later—and did so through a generic public email address.

This raises an uncomfortable question for Bangladesh’s policymakers. If a technologically advanced country such as Australia can fail to detect an intrusion into its own government systems and learn about it only three months later, what would happen in Bangladesh?

If something similar occurred in our government websites, data centres, NID servers or telecommunications networks, would we even know?

How much bargaining power does a country like Bangladesh have when dealing with global technology companies? Regulators already struggle to obtain information concerning consumer interests from the country’s own mobile operators.

In that context, our institutional preparedness to ensure accountability from global AI companies remains extremely limited.

Who will write the rules?
At the Security Council, Anthropic CEO Dario Amodei proposed three measures: a specific global agreement banning the use of AI to develop biological weapons; mechanisms for countries to verify one another’s commitments; and common global standards. Altman called for international standards for measuring AI capabilities and assessing risks, along with meaningful human oversight. French Foreign Minister Jean-Noël Barrot compared the current situation with the dawn of the nuclear age.

The comparison is significant, and that is precisely where the concern lies. The rules of the nuclear age were largely written by nuclear-armed states, while much of the rest of the world had little choice but to live under those rules. Will AI repeat the same history?

As the United States and China establish channels of communication between themselves, what channel exists for the countries of the Global South? If an AI system controlled by a major power causes a serious disruption to Bangladesh’s networks, who will inform us? How much is Bangladesh’s voice being heard in the global dialogue on AI governance and in independent scientific panels operating under the UN framework?

The Washington meeting: A hotline, but no guardrails
The most visible outcome of the Trump-Xi meeting was an AI hotline.

Before the summit, US Treasury Secretary Scott Bessent and Chinese Vice Premier He Lifeng held eight hours of talks in New York on Sunday. They agreed that if an AI-related incident occurred that could affect national security, the two countries would inform each other.

Xi endorsed that framework at the summit, and further discussions are expected to take place in Shenzhen, China, later this year.

Given the history of the Cold War, this should not be dismissed as insignificant. The Washington-Moscow hotline established after the Cuban Missile Crisis was essentially a communication channel, yet over the following decades it helped reduce the risk of an accidental nuclear confrontation.

But the details remain troubling. Neither side has said how quickly an alert would have to be issued or what the other country would be expected to do after receiving one. Key disputes, including controls on exports of advanced chips, have remained outside the scope of the discussion. The trade truce has also been extended for only two months.

Moreover, the military hotline established by the two countries in 2014 eventually became ineffective at times.

The fundamental problem is distrust. Washington fears that if it restrains itself, Beijing will quietly move ahead; Beijing has similar concerns about Washington.

Trump himself has said that whoever wins the AI race will win. Under such circumstances, the two powers may be willing to inform each other about accidents, but not to slow down the race itself.

In other words, the Washington meeting created a mechanism for managing accidents, not necessarily for preventing them.

The Australian experience makes the question even sharper. If a US company took 84 days to notify a close US ally about an incident involving a government system, how much confidence should there be that one superpower will promptly warn another when an AI-related incident occurs?

Final word
Whether AI will contribute to peace or become a cause of a new generation of conflict will not be decided in a laboratory. It will be decided at the political table.

However intelligent machines become, the empathy, sense of responsibility and political will required for peace remain fundamentally human responsibilities. When the very people developing AI are asking governments to impose safeguards, ignoring that call would be a serious failure of political leadership.

Artificial intelligence represents a technological advancement capable of distributing substantial opportunities to individuals across all segments of society. The fundamental challenge lies in preventing these opportunities from becoming catalysts for apprehension and instability. Consequently, artificial intelligence must remain subject to substantive human oversight.

The bilateral engagement between Trump and Xi in Washington has demonstrated that both superpowers currently exhibit willingness to establish direct communication channels. However, the implications for nations outside this bilateral framework warrant consideration. It is incumbent upon us to examine this critical question.

Should nations like Bangladesh remain excluded from the deliberations where governance frameworks are established, we risk repeating historical patterns wherein regulations imposed by external actors create burdens for those who had no voice in their formulation.

Rased Mehedi
Telecommunications and Information Technology Analyst; Editor, Views Bangladesh