AI catastrophe could cause damage by 2030 through cyberattacks, weapons, unemployment, environmental pressure, disinformation and concentrated power unless governments and developers strengthen safeguards, oversight and cooperation, according to Qazinform.
The assessment, Prioritization of Risks from Artificial Intelligence, draws responses from 272 specialists across 37 countries, examined 24 categories of AI catastrophe.
Under business-as-usual conditions, experts placed 18 categories above a 10% AI catastrophic risk, defined as more than one million deaths, over $100 billion in losses, or civilization-scale damage.
AI Weapons and Cyberattacks Carry Highest Risks
Dangerous AI capabilities received the highest estimated AI catastrophe probability at 21.5%.
Researchers said systems could deceive people, conduct cyber operations, support AI weapon development, and create powerful models. Harm may come from misuse, system failures, or behavior that conflicts with human goals.
AI-enabled weapons and cyberattacks followed closely at 21%. Technology could enable the first large scale cyberattack by AI, automate offensive operations and support chemical, biological, radiological, or nuclear threats.
Autonomous weapons may also lower the cost and technical barriers for states, or non-state groups, facing catastrophic AI risks.
“Cyberattacks from both state and non-state actors will be a permanent fixture of AI-related risks,” according to the report.
Power concentration carried an 18% probability, ranking among catastrophic AI risks. Developing leading AI systems requires datasets, specialized chips, computing infrastructure and investment, placing models in the hands of a few companies and governments.
Experts warned that those controlling the technology may capture most economic benefits. Competition between developers and countries carried a 16.6% AI catastrophic risk probability.
Pressure to release stronger models could make the first large scale cyberattack by AI more likely if organizations reduce safety testing. Developers investing time in safeguards may face immediate costs, while the benefits of caution are shared across society.
Is AI Contributing to Rising Unemployment?
The study revealed that AI could remove roles, weaken job quality, and direct wealth toward those controlling systems and data. Despite protections, experts placed the probability of an AI catastrophe from unemployment and inequality at 11%, showing that economic disruption may remain difficult to contain.
As for AI causing environmental damage, it remained above the 10% threshold, under a mitigation scenario, reaching 12%. Training and operating large models consume increasingly large amounts of electricity and water, while data centers depend on specialized hardware, construction, and supply chains.
AI and environmental damage could increase pressure on grids and natural resources if efficiency gains fail to match demand.
False information received a 12.8% estimate, adding to catastrophic AI risks. Generative AI can produce convincing text, images, audio and video at scale, making fabricated material harder to identify. Experts said this could support political manipulation, weaken trust in journalism and institutions, and make it harder for societies to agree on basic facts.
According to the University of Queensland, the same research identified national security, finance, and information technology among highly exposed sectors. Members of the public and AI users were considered the most vulnerable, despite having limited ability to control how powerful models are designed or deployed.
Researchers used the Delphi method, collecting three anonymous rounds of expert input from industry, academia, government, and civil society. The process was intended to reduce the influence of individual interests and produce a clearer collective judgment about severity, vulnerability, and responsibility.
Even with pragmatic safeguards, AI catastrophe remained possible across five categories above a 10% probability. Dangerous capabilities, weapons and cyberattacks, environmental harm, inequality and unemployment, and power concentration all suggest that voluntary measures and technical fixes alone may not be enough.
“These findings are a call to arms for those building and governing AI to make sure the worst outcomes do not occur,” said MIT FutureTech Director, Neil Thompson.
The authors stressed that the percentages are expert judgments, not precise forecasts, and should not be added because several risks overlap. Still, the findings place responsibility mainly on general-purpose AI developers, governments, regulators and standards bodies to strengthen testing, accountability, treaties and coordinated safeguards before 2030.
Inside Telecom provides you with an extensive list of content covering all aspects of the tech industry. Keep an eye on our Intelligent Tech sections to stay informed and up-to-date with our daily articles.
