TITLE: AI Expert Warns World May Lack Time for Safety Prep

AI Safety Concerns Escalate with Rapid Technological Advances

A leading AI safety researcher has warned that the world may not have sufficient time to address the risks posed by rapidly advancing artificial intelligence systems, potentially leading to a destabilization of global security and economies.

David Dalrymple, program director and AI safety expert at the UK's publicly funded Aria agency, emphasized the urgency in an interview. He highlighted concerns over AI systems capable of surpassing human performance across critical domains, which could undermine societal control.

"I think we should be concerned about systems that can perform all of the functions that humans perform to get things done in the world, but better," Dalrymple said. "We will be outcompeted in all of the domains that we need to be dominant in, in order to maintain control of our civilisation, society and planet."

Dalrymple pointed to a significant disconnect between public sector understanding and the pace of innovation in private AI companies. He projected that within five years, machines could handle most economically valuable tasks at higher quality and lower cost than humans, a development he described as neither science fiction nor distant.

"I would advise that things are moving really fast and we may not have time to get ahead of it from a safety perspective," he added.

Gaps in Reliability and Control Measures

Dalrymple stressed that governments cannot assume advanced AI systems will be inherently reliable. Aria, independent from direct government control, directs research funding toward safeguarding AI applications in critical infrastructure, such as energy networks.

"We can’t assume these systems are reliable. The science to do that is just not likely to materialise in time given the economic pressure," Dalrymple said. "So the next best thing that we can do, which we may be able to do in time, is to control and mitigate the downsides."

He described unchecked technological progress as a potential "destabilisation of security and economy," calling for increased technical efforts to understand and manage the behaviors of advanced AI. While acknowledging optimism among some frontier researchers that such progress could yield benefits, Dalrymple characterized the transition as high-risk, with humanity largely unprepared.

"Progress can be framed as destabilising and it could actually be good, which is what a lot of people at the frontier are hoping. I am working to try to make things go better but it’s very high risk and human civilisation is on the whole sleep walking into this transition," he said.

UK AI Security Institute's Latest Findings

Supporting Dalrymple's concerns, the UK's AI Security Institute (AISI) reported this month that capabilities of advanced AI models are improving rapidly across all domains. Performance in some areas is doubling every eight months, according to the institute.

Leading models now complete apprentice-level tasks successfully 50% of the time, a marked increase from approximately 10% the previous year. The most advanced systems can autonomously handle tasks that would require over an hour of human expert effort.

AISI also evaluated self-replication capabilities, a major safety issue due to the potential for systems to propagate uncontrollably across devices. Tests on two cutting-edge models showed success rates exceeding 60%.

However, the institute noted that worst-case scenarios remain unlikely in everyday settings. "Any attempt at self-replication was unlikely to succeed in real-world conditions," AISI stated.

Dalrymple anticipates further acceleration, predicting that by late 2026, AI could automate a full day of research and development work. This self-improvement in areas like mathematics and computer science would compound capability growth.

Broader Implications for Policy and Society

The warnings come amid growing integration of AI into daily life and critical sectors. Recent research indicates that one-third of UK citizens have turned to AI for emotional support, underscoring the technology's expanding role.

In parallel, industry pushback against AI misuse is evident. UK actors have voted to refuse digital scanning for AI purposes, aiming to protect against unauthorized use of likenesses.

Dalrymple's role at Aria involves developing safeguards for AI in high-stakes environments. The agency's independence allows it to prioritize long-term safety over short-term gains, but Dalrymple urged broader collaboration between governments and tech firms.

As AI capabilities surge, the need for proactive mitigation strategies intensifies. Without timely interventions, the expert warns, the balance between innovation and control could tip toward unintended consequences, affecting global stability.

The AISI's assessments provide a benchmark for ongoing monitoring, but Dalrymple emphasized that economic incentives may hinder the development of comprehensive reliability measures. Policymakers face the challenge of fostering innovation while embedding safety from the outset.

This situation reflects a pivotal moment in AI governance, where the pace of breakthroughs demands agile responses to preserve societal safeguards.