What catastrophes could AI cause by 2030?
Artificial intelligence could contribute to large-scale cyberattacks, weapons development, rising unemployment, environmental damage and the concentration of power among a small number of companies. These risks were identified by 272 experts from 37 countries in the report Prioritization of Risks from Artificial Intelligence, Qazinform News Agency reports.
The experts assessed 24 AI risk categories through 2030. The study defined catastrophic harm as outcomes involving, for example, more than one million deaths, financial losses exceeding $100 billion or civilization-scale damage to democratic norms, civil rights and social stability.
Under a business-as-usual scenario, experts assigned 18 of the 24 risks a probability of catastrophic harm above 10%.
Weapons and cyberattacks
Experts estimated a 21% probability that AI-enabled weapons and cyberattacks could cause catastrophic harm.
AI could facilitate the development of more effective malware, autonomous weapons and chemical, biological, radiological or nuclear threats.
“Cyberattacks from both state and non-state actors will be a permanent fixture of AI related risks,” the study says.
Dangerous AI capabilities
The highest estimated probability, 21.5%, was associated with AI systems acquiring dangerous capabilities.
These include the ability to deceive and manipulate people, conduct cyber operations, assist in weapons development and contribute to the creation of more advanced AI systems. Such capabilities could be misused by humans, but harm could also arise from system failures or behavior that conflicts with human goals.
“Dangerous capabilities emerge as an emergent property of scaling AI systems, which makes them difficult to predict and control. Even with safety measures in place, the potential for catastrophic misuse remains,” one of the experts surveyed warned.
Concentration of power
Experts placed the probability of catastrophic harm from the concentration of power and resources at 18%.
Developing advanced AI requires large amounts of data, computing capacity and investment. As a result, the most powerful systems could remain under the control of a limited number of corporations and governments.
“Power centralization is perhaps the most stubbornly persistent risk because the same entities developing AI are often best positioned to capture its benefits, creating self-reinforcing dynamics that are difficult to reverse through technical interventions alone,” the report says.
Competition between companies and states
Competitive dynamics in AI development were assigned a 16.6% probability of catastrophic harm.
The race to release increasingly powerful systems could encourage companies and governments to reduce safety testing or deploy models before their risks are fully understood.
According to the report, a developer that slows down to invest in safety bears an immediate competitive cost, while the benefits of that caution are distributed across society.
Unemployment and inequality
AI could automate jobs, reduce employment quality and widen the gap between those who control advanced technologies and the rest of society.
Even under a scenario involving pragmatic safeguards, experts estimated an 11% probability of catastrophic harm from inequality and unemployment.
“Inequality is deeply entrenched in economic and social systems. AI may exacerbate existing inequalities through automation-driven job losses in certain sectors while creating wealth for those who own and control AI systems,” one expert said.
Environmental damage
The development and operation of large AI models require increasing amounts of electricity, water, computing infrastructure and specialized hardware.
Environmental harm retained a 12% probability of catastrophic consequences even under the pragmatic-mitigation scenario.
“Heavy data infrastructure and large models will continue [to] consume energy, cause resource pressure,” one participant noted.
Loss of a shared reality
AI can generate convincing fake texts, voices, images and videos on a massive scale. This could make information harder to verify, facilitate political manipulation and undermine trust in the media and public institutions.
False information was assigned a 12.8% probability of catastrophic harm. Separately, 22.1% of participants included disinformation and influence among their three greatest concerns, while 21.6% selected the loss of consensus reality.
The authors stress that these percentages represent collective expert judgments rather than precise forecasts. They also cannot be added together because many of the risks overlap and could materialize through the same events.
Earlier, Qazinform News Agency reported that Tokyo is expanding its AI-powered matchmaking program to help more residents find potential marriage partners.
