Newsletters
Artificial intelligence is advancing so quickly that it could greatly increase the abilities of criminals, terrorist groups, and hostile states, according to a new report from the University of Cambridge's Center for the Study of Existential Risk (CSER).
The report says that more advanced AI systems could make cyberattacks cheaper, faster, and easier for more people to carry out. It also warns that AI could speed up the spread of online disinformation, fraud, biological and chemical threats, and the use of autonomous weapons in the military.
The Cambridge researchers call AI a classic "dual-use" technology, meaning it can bring big benefits but also create new risks for abuse. The report emphasizes that the real challenge is ensuring safety measures keep pace with AI's rapid progress, not the technology itself.
These findings come as top AI companies compete to create a new generation of "frontier" AI models. Companies like OpenAI, Anthropic, Google DeepMind, Meta, and xAI have launched advanced language models that can reason, write software, conduct scientific research, and handle complex tasks with minimal human assistance.
Unlike older AI systems that mostly generated text, today's frontier models can write and fix computer code, analyze large amounts of data, browse the internet, use digital tools, plan complex tasks, and work on long projects.
Britain's AI Security Institute found that AI is improving at cybersecurity, with some capabilities doubling about every eight months. Recent frontier models have already demonstrated expert-level performance on certain cybersecurity tasks.
The Cambridge researchers say cybercrime may be the most urgent risk. Advanced AI systems can automatically identify software weaknesses, create convincing phishing emails in many languages, write malicious code, analyze stolen data, and help orchestrate more complex cyberattacks.
Earlier this week, the Five Eyes intelligence alliance, which includes the United States, Britain, Canada, Australia, and New Zealand, warned that AI models capable of enabling highly advanced cyberattacks against governments, businesses, and critical infrastructure could emerge within months.
The report also highlights the fast spread of AI-generated misinformation as a major concern. As AI-made images, videos, and voices become more realistic, hostile governments, criminal groups, and influence campaigns could use fake content on a huge scale to sway elections, increase political divides, or spread false stories faster than fact-checkers can keep up.
The report also points out the rise of autonomous AI agents capable of performing long chains of actions with little human supervision. Unlike regular chatbots, these systems can work with software independently, make decisions, complete tasks, and use multiple digital tools. This reduces the chances for people to step in if something goes wrong or if the systems are misused on purpose.
Researchers also warn that AI could play a bigger role in research involving biological and chemical agents. While today's models are limited and have strong safeguards, future systems might have scientific capabilities that require stricter oversight to prevent abuse while still enabling genuine medical and pharmaceutical research.
Fast progress in AI-powered drones, surveillance, and autonomous weapons is already changing modern warfare. Security experts warn that AI technology sold to the public could be used not just by governments but also by other groups, making future conflicts faster, larger, and more precise.
Despite the rising risks, the report does not say that AI development should stop. Instead, the Cambridge researchers suggest better teamwork between governments, AI developers, and cybersecurity experts, more openness about advanced AI systems, more thorough safety testing before launch, and closer international coordination on AI governance.
These findings align with the broader International AI Safety Report 2026, which states that AI is advancing faster than many current safety measures can keep pace with. It lists malicious use, technical failures, and systemic risks as the technology's three most pressing challenges.
The Cambridge researchers say the world has only a short time to get ready before AI becomes much more powerful. "The same technology that can accelerate scientific discovery can also accelerate cybercrime," the report says.
It adds that stopping misuse will require ongoing monitoring, international collaboration, and security measures that keep pace with AI progress.
Set as preferred source
Set as preferred source
1 min read
1 min read
