MIT Researchers Warn of Major Disaster, Millions of Deaths in 5 Years

Jakarta, CNBC Indonesia –The development of artificial intelligence (AI) technology has created a major dilemma for entrepreneurs seeking to capitalize. On the one hand, they are constantly competing to win the AI ​​race to dominate the market.

On the other hand, AI has had a negative impact on various sectors. The spread of disinformation, water and electricity crises, and waves of layoffs are some of the serious problems that are already clearly visible.

There are many other impacts that are also significant, but less frequently publicized. For example, increasingly sophisticated cyberattacks, easier development of advanced military weapons, and other indirect impacts.

A new study from MIT FutureTech and the University of Queensland offers a way to map this risk landscape. In the report, “Prioritization of Risks From Artificial Intelligence,” a research team including Neil Thompson, principal research scientist at the MIT Sloan School of Management, asked 272 international AI experts to evaluate 24 AI risks based on their likelihood of occurrence and severity.

They also noted which sectors and actors are most vulnerable, and who should be responsible for addressing these risks.

Experts have identified five key risks with the potential to have the most severe impacts in the next five years. These are the emergence of dangerous capabilities, competitive pressures, cyber weapons and attacks, the concentration of power, and misinformation.

The information and financial sectors are considered highly vulnerable, and the people and organizations most vulnerable to these risks are often ill-positioned to address them.

“There are a lot of risks with AI,” said Peter Slattery, a research scientist at MIT FutureTech and one of the study’s co-authors.

“One of the key points behind this research is trying to understand who needs to do what differently, and in what form of coordination,” he explained.

The Most Severe Risks

This study used the Delphi method, a structured research process that gathers expert assessments through several stages to reach consensus and more clearly identify points of difference. The researchers used these expert assessments to evaluate risks over a five-year time horizon (2025 to 2030) under two scenarios:

1. Business as usual (conditions without policy changes): Organizations and governments continue their current approaches.

2. Pragmatic mitigation: Organizations and governments undertake cost-effective efforts to mitigate AI risks.

Under a business-as-usual scenario, experts assess that 18 of the 24 AI risk domains have at least a 10% probability of triggering a catastrophic impact within the next five years. A catastrophic outcome is defined as a loss that includes the potential for more than 1 million deaths, financial losses exceeding US$100 billion (Rp1.8 trillion), or intangible damage equivalent to civilization-scale.

Pragmatic mitigation does lower these estimates, but does not eliminate the risk entirely. Even with efficient mitigation efforts, experts assess that five domains still have at least a 10% probability of triggering catastrophic impacts:

. AI systems with dangerous capabilities – 12% probability

. AI weapons, cyberattacks, or other mass-harm capabilities – 12% probability

. Environmental destruction – 12% probability

. Social inequality and unemployment – 11% probability

. Concentration of power and unfair distribution of AI benefits – 11% probability

AI-powered weapons and cyberattacks rank high because AI is well-suited to coding, pattern recognition, and information synthesis, Slattery explained. Much of modern infrastructure relies on software, which provides a significant attack surface for systems that can help hackers identify vulnerabilities, generate code, or accelerate offensive cyber actions.

“AI is highly capable of attacking these aspects of society and causing harm,” Slattery said. “Coding and hacking are some of the areas where we’re seeing the fastest growth in AI capabilities,” he said.

Key Risk Dynamics

. Dangerous Capabilities: The scope is broader. While malicious actors can use AI systems for harm, advances in these systems can also facilitate difficult and dangerous tasks, such as mass persuasion, surveillance, deepfake creation, and even aiding in the development of chemical or biological weapons. “These are things that could be done before, but now they can be done much more easily,” Slattery said.

. Competitive Pressures: It’s not one form of AI use that’s dangerous, but rather the conditions that can exacerbate other risks. When companies or nations believe AI will provide a significant economic or strategic advantage, they are encouraged to move quickly, ignore limitations, or underinvest in safety. “These are instrumental risks that create other risks,” Slattery said.

Most Affected Sectors and Responsible Parties

This study identifies three sectors that experts believe will be most vulnerable to AI risks in the next five years: information, national security, and finance. Leaders in these sectors must understand that the risks they face are specific to their respective characteristics:

. Information Sector: Risks are closely related to the flow of content and data, including misinformation, disinformation, loss of privacy, manipulation, and the erosion of trust in what the public sees and reads.

. National Security: Top concerns include cyberattacks, weapons development, massive surveillance, and the use of advanced AI systems by hostile actors.

Finance: AI can amplify fraud, cyber risks, market manipulation, privacy violations, and systemic failures that have broad economic impacts.

Across these sectors, AI is making existing risks more scalable. People who previously couldn’t hack can now do hacking-related tasks, while those with expertise can work faster, better, and more easily, Slattery explains.

Experts state that AI developers and governance institutions (governments and regulators) bear the primary responsibility for addressing AI risks. However, users of AI systems and the communities affected by them are the most vulnerable. This misalignment between those responsible and those most vulnerable creates a gap in action.

What World Leaders Should Do

Slattery emphasized that the research doesn’t aim to predict the future with certainty, but rather to help leaders focus on AI risks that are considered serious and plausible in the short term. “We’re not saying these things will definitely happen,” he said.

“We’re saying these are the things that experts say need to be addressed now,” he said.

For business leaders, these key risks raise two pressing questions:

1. What can today’s increasingly sophisticated AI systems do?

2. Are competitive pressures pushing organizations to implement AI faster than their governance can keep up with it?

The concern is that companies, industry sectors, and governments feel pressured to move quickly even when the risks are not yet fully understood.

That’s what makes these findings so valuable for business leaders, especially considering that many organizations still treat AI risks as merely compliance issues or distant future threats. Leaders must evaluate AI at the business process level: figuring out where AI can deliver return on investment (ROI), where it can transform the broader ecosystem, and whether it could disrupt the products or services their organization provides.

Executives must begin to recognize that AI risks are not business-as-usual threats. Slattery cautions against abandoning what’s working, but leaders must treat AI as a new paradigm that could replace many human tasks, introduce new vulnerabilities, and create “opportunities and vulnerabilities at the ecosystem and societal levels.”

These adjustments also can’t be a one-time effort. “This must be a continuous and ongoing process starting now, because AI is moving so rapidly that organizations need to be much more responsive to this technology than they have been in the past,” Slattery said.

The key message is that organizations don’t need to wait for perfect forecasts or formal regulations before acting. They can start by focusing on the most severe and likely potential harms and incorporating AI risks into existing governance discussions around cybersecurity, privacy, safety, and business continuity.

The report “Prioritization of Risks From Artificial Intelligence” was written by 188 co-authors, with a core research team consisting of Alexander K. Saeri, Jess Graham, Michael Noetel, Peter Slattery, and Neil Thompson.

This research is part of the MIT AI Risk Initiative by the MIT FutureTech lab. MIT FutureTech is an interdisciplinary group affiliated with the MIT Sloan School of Management and the MIT Computer Science and Artificial Intelligence Lab that studies the economic and technical foundations of computing advances.

The MIT AI Risk Initiative aims to provide authoritative data and frameworks to help identify, prioritize, and manage risks from AI. This includes the AI ​​Risk Repository, an updated database of over 1,600 AI risks grouped by cause and risk domain, which is used by policymakers, technologists, and organizations to design AI governance.

SOURCE : CNBC INDONESIA