A new study by researchers from MIT FutureTech and the University of Queensland identified five AI risks that experts rated among those with the highest expected severity over the next five years.

The study, titled “Prioritization of Risks from Artificial Intelligence,” involved 272 international AI experts and assessed 24 AI risks based on their likelihood and severity, as well as vulnerable sectors and groups and responsibility for addressing them.

Across most risks, experts identified the information, finance, and national security sectors as the most vulnerable. In the information sector, major concerns include misinformation, disinformation, privacy loss, manipulation and declining trust in digital content.

In national security, concerns include cyberattacks, weapons development and surveillance. In finance, concerns include fraud, cyberattacks, market manipulation and privacy breaches.

A new study asked 272 experts to evaluate AI risks and the sectors that are most vulnerable over the next five years.

Learn more: https://t.co/0Eq7ES0QiY pic.twitter.com/k4GVcx2bgS

— MIT Sloan School of Management (@MITSloan) August 11, 2026

The five risks with the highest expected severity were dangerous capabilities, competitive dynamics, weapons and cyberattacks, power centralization, and false or misleading information.

When asked separately to select the three AI risks they were most concerned about causing harm over the next five years, experts most commonly chose weapons and cyberattacks (26.8%), power centralization (23.5%), disinformation and influence (22.1%), loss of consensus reality (21.6%), and dangerous capabilities (21.6%). The study noted that these concerns did not completely match the severity rankings.

The study assessed the risks over a five-year horizon, from 2025 to 2030, under two scenarios: Business as Usual, in which organizations and governments continue their current paths, and Pragmatic Mitigations, in which they make cost-effective efforts to reduce AI risks.

Under the business-as-usual scenario, experts estimated that 18 of the 24 risk areas had at least a 10% probability of causing catastrophic harm over the next five years. The study defined catastrophic harm as events that could cause more than 1 million deaths, more than $100 billion in financial losses, or comparable civilizational-scale intangible harms.

Even under the Pragmatic Mitigations scenario, five risks were still judged to have at least a 10% probability of catastrophic outcomes. These were dangerous capabilities (12%), weapons and cyberattacks (12%), environmental harm (12%), inequality and unemployment (11%), and power centralization (11%). All 24 risks were judged to have more than a 5% probability of catastrophic outcomes.

Peter Slattery, a research scientist with MIT FutureTech and one of the study’s co-authors, said the research aims to figure out “who needs to do what differently, and in what sort of coordination.”

The study stated that, according to experts, AI developers and governance actors, such as governments and regulators, have the primary responsibility for addressing AI risks. But AI system users and the stakeholders affected by these systems are most vulnerable to those risks. Because those most responsible for addressing the risks are not those most vulnerable to them, the study found misaligned incentives for taking action.

Slattery said the research is not intended to predict the future with certainty, but rather to help leaders focus on AI risks that experts consider serious and plausible in the near term. “We’re not saying these things are definitely going to happen,” he said. “We’re saying these are the things that experts think are worth paying attention to now.”

According to Slattery, leaders should build on existing approaches while treating AI as a new paradigm that could replace many human tasks, introduce new vulnerabilities, and create “ecosystem-wide, society-wide opportunities and vulnerabilities.”

The paper was co-authored by 188 researchers, with a core research team of Alexander K. Saeri, Jess Graham, Michael Noetel, Peter Slattery, and Neil Thompson.

Also Read: What Happens When AI is Induced to Assert its Own Consciousness? Inside Google’s Latest Paper