BOSTON, August 10, 2026: Artificial intelligence technology is evolving at an incredible pace with emerging prospects in science, business, medicine, and even our daily lives. However, alongside the evolution of AI, there appears an increasing attention to the challenge that is posed by the problem of rapid technological advancement surpassing the regulatory systems that are needed to manage AI risks.
Recently, an international expert study has provided the broadest evaluation to date of that challenge. Scientists from MIT FutureTech and the University of Queensland invited 272 experts from various academic, industrial, governmental and other backgrounds to evaluate the risk of 24 types of hazards connected with artificial intelligence.
That three-round international study does not limit itself to the well-known risks of AI. It invites experts to analyze not just the usual issues, such as job loss, bias, or misinformation, but to estimate the likelihood of those risks and severity of possible consequences, to find out the industries and populations vulnerable to them, and to define the responsible parties to address the risks.
Potentially Dangerous AI Capabilities Identified as Major Threat
Among the 24 areas considered, potentially dangerous capabilities of AI were seen as the most pressing one.
The definition includes cases when more advanced AI could be used to perform activities that could cause substantial damage. With increasing levels of capabilities in reasoning, planning, programming, science, and tool operation, there comes a point when the AI systems become capable of activities that require highly specialized knowledge from humans.
Obviously, all this provides immense advantages. Capabilities that can aid scientists to understand complicated issues in their field or make their job easier through automation could also serve as an opportunity for exploitation.
Thus, the threat posed by potentially dangerous capabilities of AI is not just in its growing intelligence but also in expanding capabilities with possible negative consequences.
There is a Risk That the Race to Lead in AI Will Lead to Further Risk
The competitiveness of the situation was a third big worry.
The AI field globally is very competitive. Tech firms have put a lot of money into infrastructure, people, data, model training, and their applications. Meanwhile, governments are starting to view advanced AI as an economically and strategically important technology.
Healthy competition can be good. It can inspire innovation, lower costs, improve products, and speed up the process of developing valuable technologies.
On the other hand, competition can be a source of risk management difficulties.
Firms might feel compelled to get new systems out on the market since the competition is moving ahead of them. Investors want quick growth. The clients require more and more sophisticated capabilities from the product. The researchers need to prove progress even if some risks are not fully studied.
It means that there is a systemic issue described by the specialists. It is not possible for any company to stop despite its recognition of safety importance.
Weapons and Cyber Attacks Create Security Risks
Weapons and cyberattacks also belong to the five fields that have the greatest possibility to inflict serious harm.
Artificial intelligence is already employed in cybersecurity and cyberattacks detection and protection. Meanwhile, the growing power of artificial intelligence may be potentially exploited by attackers to enhance some types of harmful activities.
The main risk that emerges here is the ability of the technology to enhance existing possibilities.
Various cyber attacks and operations that previously required much knowledge can be easily expanded due to the growing programming skills of AI.
It doesn’t mean that every AI system turns into a cyber attack tool automatically. It means, however, that governments and cybersecurity experts should think carefully about the possible influence of the technology on the balance between attackers and defenders.
Military applications add yet another issue.
Artificial intelligence can facilitate intelligence analysis, logistics, planning and many other activities related to defense. As military establishments develop AI systems, there appear a number of hard questions about human control, accountability and consequences of employing such a technology.
Concentration of Power May Affect the Future of AI
Another critical problem highlighted by experts is the concentration of AI capacities and benefits.
The creation of the most advanced AI systems requires huge amounts of computing capacity, highly-developed infrastructure, vast amounts of data, skilled researchers, and substantial investment.
All of the mentioned factors can present certain barriers to entry into the market.
If advanced AI technologies start to be developed only by a few companies or organizations, these organizations may have an impact on technologies that will affect large segments of society.
It is not just about the ownership of AI systems.
It may also come from such factors as computing capacity, data, distribution channels, cloud computing, platforms, and applications. Companies that hold important positions in several areas of the AI industry may exercise substantial power over the evolution of the technology.
Misleading or False Information Could Be Harder to Spot
Misleading or false information is one of the most commonly known risks associated with AI technology, but the study puts it in the category with the highest risks of serious consequences.
With generative AI, one could generate real text, images, audio, and video at an enormous scale not seen before.
A person trying to spread misleading information could use automation to do so on a large scale. This would become much harder when people find it hard to distinguish between genuine content and that produced by AI.
This risk could have far-reaching effects.
One area where this could happen is elections. It could be used to impersonate public figures and spread misinformation.
Catastrophic Scenarios Still Pose a Major Threat
One of the most notable aspects of the research is its evaluation of risks associated with catastrophes.
If current trends continue, 18 out of 24 categories of AI risks were estimated to be over 10 percent likely to cause a catastrophic scenario within the period from 2025 to 2030.
The research applies a rather broad understanding of what a catastrophe is.
Specifically, any scenario involving more than one million fatalities, more than $100 billion in economic losses, or civilization-wide damage to intangible values is considered to be a catastrophe.
These figures do not imply that such disasters are expected to occur.
Probability estimates provided by experts are measures of uncertainty, not forecasts of definite future events.
Nonetheless, the data show the gravity with which specialists approach the threat of AI failure or misusage.
Furthermore, the research makes a comparison between different scenarios.
In the case of a pragmatic approach to risk reduction, the following categories still had more than 10 percent probability of causing a catastrophe: dangerous capabilities, attacks and cyber attacks, environment, inequality and unemployment, and power centralization.
AI Risks Will Not Affect Everybody the Same Way
Yet another finding that stands out relates to the link between vulnerability and responsibility.
The experts regarded AI users and the general population as the two parties most exposed to AI risks.
It is obvious why.
Regular users rarely have any influence over training, datasets, protections, and infrastructure that goes into an AI model but might experience negative effects due to the AI providing inaccurate data, leaking personal data, making biased decisions, or becoming unavailable.
Companies may find themselves in a similar situation.
The reliance on an AI-based service of a company that cannot control the technology itself might cause problems due to any actions taken by the provider – even if there were no faults committed by the customer.
For that reason, the experts allocated major responsibilities to both AI system developers and governance parties such as governments, regulatory bodies, and standard organizations.
Developers Play a Crucial Part
AI developers are some of the organizations in the best position to know the technical strengths and limitations of their systems.
This means developers play a particularly crucial part in risk management.
Before putting powerful models into circulation, the companies could carry out tests aimed at uncovering any risky features of the system, including security risks, reliability issues, and misuse.
They could set up monitoring mechanisms and put in place policies for dealing with any new risks that are identified.
However, the results from this study show that it might not be enough to leave risk management completely to companies.
The commercial companies have legitimate reasons for making money and competing. These are not always aligned with wider social interests.
Balancing Act for Governments
Governments must manage a number of different aims.
They wish to foster innovation and growth but also to protect their citizens from potential harms.
They wish to compete in the development of AI technologies but also to cooperate in the management of cross-border risks.
The problem is especially difficult since any regulation of AI could impact investments.
Insufficient regulation would be ineffective, while poor or overly restrictive regulation could stifle innovation or relocation of development efforts.
It is thus not simply a matter of making more laws and regulations, but rather one of assessing what risks merit regulation and what safeguards are feasible and how these policies will be enforced.
In addition, AI may evolve more rapidly than traditional legislative process allows.
Governments might therefore require frameworks that are adaptable to new evidence.
Organizations Must Think More Broadly About Risks of AI
From the perspective of organizations adopting AI, the research findings imply that they should not restrict their risk management approaches to cybersecurity and regulatory compliance issues.
There are a number of AI risks that organizations may encounter.
The organization can suffer from brand dilution and misinformation, financial losses due to AI-supported fraud, exposure of personal data, biased outcomes, workforce changes.
Finally, there is a risk of dependence.
An organization that incorporates an AI system can become dependent on the external service provider whose terms of service, pricing, model behavior, and availability of the product can change in unpredictable ways, and the customer has little or no influence on it.
Therefore, organizations must ask themselves such questions as:
Whose is the AI system?
What if something goes wrong – what will be the consequences?
How will sensitive data be protected?
Can humans review the decision made by the AI?
What if the provider of the system becomes inaccessible?
What is the response time for a security issue?
All these questions become more and more relevant as AI technologies move beyond proof-of-concept projects and into business processes.
Inequality and Employment Are Still Key Issues
Even though job loss is not one of the top five high severity categories found through the research, inequality and unemployment issues were still prominent enough to make it on the list of those categories with high risk assessments when the mitigated scenario is considered.
AI will bring productivity increases, but the fact is that not everybody will necessarily share in them.
There may be some who will find their tasks become easier and more productive. Others will find some of what they do can be automated.
These impacts may differ greatly across different industries and occupations.
Those whose tasks are very repetitive may be more affected. For professionals, a different type of change will come from AI taking over some of the research, analysis, or administrative work.
New jobs will also be generated.
Traditionally, the evolution of technology has seen the rise and fall of whole industries and occupations at once. The problem is that this process is not always evenly distributed.
Those will have to learn new skills, while education and training systems will struggle to adapt.
Five Risks That Need Immediate Attention
The report’s findings can be stated in five areas of particular importance.
- Hazardous Capabilities
The fast growth of AI capabilities might increase risks of abuse or unintended behavior. Evaluation of the safety should be able to catch up with advanced developments.
- Competitive Dynamics
Interorganizational and interstate competition might accelerate development. Standards and governance systems may be necessary to mitigate the risk of competition undermining safety incentives.
- Weapons and Cyberattacks
AI may facilitate existing cybersecurity and military capabilities. The security organizations should be ready not only to use it for defensive purposes but also to consider misuse possibilities.
- Power Concentration
Advanced AI technologies can be concentrated in the hands of those organizations that have significant resources such as capital, computing power, data and technical expertise. It might affect competition and opportunities.
- False Information
The fast creation of false information by means of synthetic media becomes cheaper and easier. Verification tools and literacy measures become more important.
Planning for an Uncertain AI Future
The future of artificial intelligence will be uncertain.
Some of the forecasts about the technology will turn out to be incorrect. Risks that have not yet been thought of could arise. Concerns about certain risks will fade as better technical safeguards are introduced.
The strength of an analysis such as this is thus not that it gives a definitive outlook.
The strength is that it enables decision makers to allocate priorities.
Given the 24 types of risk identified in the analysis, governments and companies will not be able to give the same priority to all scenarios. They are bound to differ greatly in the degree of impact they could have on us.
According to the experts’ assessment, capabilities that can be dangerous, pressures from competition, weapons and cyber attacks, and concentrations of power and misinformation stand out.
At the same time, there are other risks, too.
The Coming Five Years May Prove Critical
The years 2025 to 2030 may turn out to be critical for AI governance.
During this period, the integration of AI technologies into commercial ventures, governments and personal uses will grow significantly.
The decisions taken today might prove crucial for the future development of the technology.
As long as safety protocols, transparency standards and governance frameworks develop in tandem with AI capabilities, most risks could be minimized.
If they do not evolve in line with advances in AI, societies would have to address issues that had become too large by then.
This is why the researchers’ analysis should be seen more as a warning of the need to be ready than as a prediction of an impending disaster.
The idea is not to anticipate the worst case scenario.
The idea is to make sure that the world will be ready in case powerful AI technologies pose challenges to which the existing institutions have no answers.
Conclusion
The analysis of 272 international experts gives a broad perspective on the possible dangers that may come with the fast growth of artificial intelligence.
The study’s results go beyond the usual topics of bias, misinformation and job loss and draw attention to a larger range of dangers associated with dangerous features, competition, cyberattacks, weaponization, monopoly and information integrity.
The study also calls attention to the possible scale of the danger. Within the framework of business-as-usual scenarios used by the researchers, experts estimated the probability of catastrophic consequences for many out of 24 examined categories. Even taking into account realistic mitigation efforts, several categories remained over the 10% level.
Those numbers do not mean that catastrophic consequences will occur inevitably.
They prove only the need to prepare for those risks.
Development of AI will continue. The companies will invest their resources. Competitions and cooperation between governments will continue. Researchers will develop better systems, and people will continue finding new uses for them.
The only thing that remains to be seen is how the safety, accountability and governance of those technologies will catch up.
