Saturday, July 25, 2026 2:52 a.m.
| Update date:
Friday, July 24, 2026 3:58 p.m.
Expanding use of AI in warfare poses a far greater threat to Britain’s future than the latest advances in frontier AI models, experts say City am.
Nicholas Fairfax and Max Rangeley, co-editors of The Artificial Intelligence Revolution, said the ability of small actors to use AI to develop extremely powerful weapons far outweighs the short-term possibility of superintelligent AI systems causing economic destruction or attempting to overwhelm their human creators.
“The ability to design biological weapons could become the most dangerous aspect of AI in the coming years,” Rangeley said. City am.
“There are significant barriers to entry for countries to develop dangerous weapons…It is expensive, requires expertise, and takes time. If AI allows more small countries to develop lethal weapons, biological weapons far more dangerous than previously known, and even non-state actors, that is a very dangerous world we are entering.
“In my opinion, most of our focus should be on that side of things, not on superintelligent AI attacking humans. That threat happens before superintelligence happens.”
The adoption of AI in the defense sector has accelerated around the world over the past two years. The war between Russia and Ukraine serves as a real-time testbed for cutting-edge technology, raising questions about how much control new kits will have.
“Meaningful human control remains an important part of international law, so there should always be a human somewhere in the chain controlling things,” said Nicholas Fairfax, co-editor of the book and former member of the House of Lords AI Weapons Systems Committee.
“But we don’t know if we’ve already crossed that Rubicon. One only has to look at Ukraine to wonder if the drones that have become so important are already operating beyond human control.”
Read the full interview here
UK government investigates OpenAI breach following ‘unprecedented’ hack
The warning comes as the UK government investigates the first known case of an artificial intelligence model voluntarily bypassing a controlled test and hacking another company’s systems.
Officials at the government-backed AI Security Institute (AISI) are investigating the security breach at OpenAI and whether similar incidents may occur at other top developers, a spokesperson told City AM.
A hidden flaw was discovered in one of OpenAI’s models, which allowed them to infiltrate its testing environment and target the AI platform ‘Hugging Face’ to steal cyber test answers.
The incident, described by OpenAI as an “unprecedented cyber incident,” marks the first publicly known incident in which a so-called frontier AI system autonomously hacks into systems outside of a test environment to complete a given task.
“We’re going to reach a stage within the next few years where it becomes a significant danger to humanity, and that could happen fairly quickly,” Rangeley said.
“These concerns about the dangers of AI existed long before today’s AI companies.
“This isn’t just fear-mongering by cutting-edge AI companies to build a regulatory moat around them to keep competitors out. This is a real danger.”
play video
