Airmic 2026: Businesses need to focus on delivering what AI cannot

Organisations face one of their greatest risk management challenges as artificial intelligence advances faster than their ability to identify, understand and control the risks.

At the Airmic annual conference in Birmingham, delegates heard that AI has become a weapon in the fight for global supremacy, leaving businesses at risk of being caught in the crossfire.

Dr Keith Dear, CEO and founder of intelligence firm Cassi, told the conference that organisations need to manage AI carefully and fully understand its outcomes.

“I think since 2014 AI is resulting in the most profound revolution in human history,” he explained.

Dear said the Mythos system highlighted the scale of the risks AI could pose.

“The release by Anthropic of its large language model Mythos, was delayed because it was viewed as a threat to the future of humanity,” he continued. “The system has the ability to identify bugs and vulnerabilities in systems which can be positive but the ability to identify those weaknesses would be extremely dangerous in the wrong hands.”

He explained that the system had identified vulnerabilities in a wide range of live systems within weeks.

“For example, when you see naval vessels, they will be running a very old version of software,” Dear added. “They do so because there have been years of work to upgrade those systems and identify bugs and vulnerabilities. However, that established software have also seen new bugs identified, by the new AI systems.”

Global approaches to AI

Dear said nation states are taking different approaches to AI, with the US focused on scaling, China pursuing open-source models and Europe taking a more restrictive regulatory path.

“We have seen a huge increase in investment in AI in the United States,” Dear said. “They are looking at rapid scaling. When you increase the data and computing capabilities it delivers better performance.

“At present the computing capacity for AI is doubling every seven to nine months. The US is investing in data centres which are being built rather than spoken about in the UK and Europe.

“In terms of China they are not willing to pour huge amounts of capital expenditure, but rather build open-source models. By various means they are content to remain three to six months behind the USA, so if the US slows, they can catch up.

“Europe, including the UK, is regulating itself out of existence. They view the growth as a bubble.”

Dear told risk managers that new AI systems are already delivering significant intelligence gains.

“If a country can outperform its peers in terms of intelligence, they will be hugely ahead.”

However, Dear added that the Trump administration has put significant restrictions on the use of Mythos, demanding that it can test any upgrades and restricting its use to US firms rather than sharing it internationally.

“Intelligence demands energy and Trump is effectively using war time powers to roll put the US grid. If the intelligence is unlimited so will be the demand for energy and that brings with it risks.”

Where human judgement still counts

Dear said the human brain struggles with exponential change, making it harder to imagine how quickly AI could reshape jobs, markets and risk.

“The human brain struggles with exponentials. We cannot think the unthinkable.

“It could turn against us. It could be asked to do something, and it is misaligned with serious outcomes. I think these two scenarios are unlikely however, a jobs revolution could happen and could happen quickly.”

Dear said the work humans do will almost inevitably change, although machines still have limitations.

“AI cannot tell you what you should want,” Dear explained. “AI does not want anything it does not have any drive.

“Humans should figure out exactly what they want and what trade-offs they are willing to accept.”

For businesses, Dear said the priority is to identify where AI has limits and where human judgement remains essential.

“You have to be clear as to what your AI cannot do,” Dear explained. “Most industries and organisations need to have benchmarks around what AI and machines cannot do, and those things are what you build your business around.”