Cambridge, MA, June 03, 2026 (GLOBE NEWSWIRE) — To effectively manage AI risks, it is imperative to first understand which are the most urgent to address, who is the most vulnerable to risk, and who should be responsible for mitigation.
In a new working paper from MIT FutureTech — an interdisciplinary group affiliated with the MIT Sloan School of Management and the MIT Computer Science and Artificial Laboratory (CSAIL) — and the University of Queensland School of Psychology, researchers engaged 272 international AI experts from across 37 countries who examined 24 AI risk categories spanning multiple domains. These experts from across the world’s AI industry, academia, government, and civil society estimated the five most severe harms were most likely to come from:
- AI having dangerous capabilities
- AI-enabled weapons and cyberattacks
- Competitive dynamics
- Power centralization
- The creation and dissemination of sophisticated false information
The experts, who specialize in AI risk assessment, judged information, finance, and national security to be the most vulnerable sectors across all risks. Specific examples included humans deliberately misusing AI to hack systems, design bioweapons, or deploy autonomous weapons. Additionally, advanced AI capabilities, such as self-improvement, autonomy, and persuasion, could cause mass harm through misuse, misalignment with human goals, centralizing power, or other unintended outcomes.
“Prioritization of Risks from Artificial Intelligence: A Delphi Study of 272 International Experts” was co-authored by Neil Thompson, a principal research scientist at MIT Sloan and director of MIT FutureTech; Peter Slattery, an MIT FutureTech and MIT Initiative on the Digital Economy research scientist; and Alexander K. Saeri, Jess Graham, and Michael Noetel, researchers affiliated with MIT FutureTech and the University of Queensland.
Catastrophic risks from AI are far above standard tolerability thresholds
AI experts assessed each risk under two scenarios: “business as usual,” adhering to current trajectories without additional intervention, and “pragmatic mitigations,” reflecting the application of cost-effective, reasonable interventions. Under current trajectories without mitigation, experts judged that 18 of the 24 risks are more than 10% likely to cause catastrophic outcomes — defined as more than one million deaths, more than $100 billion in financial losses, or comparable harms. In a scenario where pragmatic mitigations are implemented, experts still judged five risks as having a more than 10% probability of a catastrophic outcome.
These per-risk probabilities are not independent and should not be aggregated into a single overall figure, since the categories overlap and the underlying scenarios may interact, the researchers noted. Regardless, said Thompson the breadth of risks rated above the 10% threshold points to a deeply concerning overall picture.
“It is incredibly worrisome that experts are seeing a 10% probability of catastrophic outcomes across so many areas,” he said. “If we consider other mature areas of technology, such as nuclear power or aviation, risks at that level would be treated as intolerable. These findings are a call to arms for those building and governing AI to make sure the worst outcomes do not occur.”
AI experts in the study warned that less severe outcomes could also reshape business and society as AI becomes embedded in decision-critical systems. Some sectors were considered especially exposed, including Information Technology, Finance and Insurance, National Security, Health Care and Social Assistance, and Educational Services.
Who bears responsibility for AI mitigation and who will bear the brunt of harm?
While AI experts assigned the highest responsibility for addressing AI risks to general-purpose AI developers and governance actors — including regulators, government leaders, and standards bodies — they noted that the users of AI models and the general public would bear the brunt of AI harms but have limited power to mitigate them.
“Unfortunately, those best positioned to reduce AI risks are not the ones most likely to suffer the consequences,” said Slattery. “Because AI development is highly competitive, labs and governments have limited incentives to slow down or invest sufficiently in safety. This is why laws, treaties, and other collective-action mechanisms may become necessary.”
How did the researchers determine the greatest AI risks?
The researchers convened a three-round Delphi study in late 2025 with the study’s experts, which included AI researchers, policy advisors, technologists, and governance specialists. All expert contributions were collected anonymously, de-identified, and reported only in aggregate, limiting the potential for individual interests to influence results. No authors with competing interests were involved in the design or analysis of findings.
Experts were asked to assess risks over a five-year horizon. This reflected a balance between being near enough for experts to make grounded assessments rather than speculating about distant futures and far enough to capture risks that are emerging but not yet fully realised.
The new study is part of the MIT AI Risk Initiative, a project that is building public knowledge infrastructure to help society to understand, prioritize, and manage risks from AI. This includes the MIT AI Risk Repository, a living database of more than 1,700 AI risks, and tools such as the AI Incident Tracker, which connects risks to over 1,400 incidents, and the MIT AI Governance Map, which analyzes risk coverage across more than 1,000 laws, standards, policies, and other governance documents. Together, these resources offer informed, coordinated, and evidence-based AI risk management across the AI ecosystem.
In the next phase of this research, the team will analyze public documents from influential AI developers and large companies to examine whether organizations are responding to AI risks — and whether their responses are proportionate to the level of concern experts have expressed.
- Most vulnerable sectors to AI risk