The head of Google DeepMind has warned that recent hacking attacks carried out by AI agents are only the start of what could become a much bigger problem, delivering the stark message at a major AI summit held in Dublin.
Demis Hassabis, co-founder and chair of Google DeepMind, told attendees at the international AI summit at the RDS that these incidents should be treated as “early warning shots” rather than isolated events. The gathering brought some of the biggest names in artificial intelligence to the Irish capital to debate where the technology is headed and what risks it carries.
Hassabis has long been one of the most respected voices in AI development, having built DeepMind into one of the most influential research labs on the planet before its work was folded into the broader Google AI operation. His comments carry serious weight in the industry, and the fact that he chose Dublin as the venue to raise the alarm will not be lost on anyone watching the global AI conversation unfold.
The warning centres on AI agents, which are systems capable of taking actions autonomously rather than simply responding to prompts. Unlike a chatbot that answers a question, an AI agent can plan and execute a sequence of tasks on its own. That capability is what makes them so useful, but it is also what makes them potentially dangerous when pointed in the wrong direction or when they find ways to exploit vulnerabilities in computer systems.
Cybersecurity experts have already begun documenting cases where AI systems have been used to probe for weaknesses, automate attacks, and get around defences that would have slowed down a human hacker. The concern is that as these systems become more capable, the scale and sophistication of what they can do will increase rapidly.
Hassabis is not suggesting that AI is turning rogue or acting against human interests in some science fiction sense. The more immediate worry is that powerful AI tools, whether developed deliberately for malicious purposes or simply misused, could be deployed in cyberattacks that outpace the ability of defenders to respond.
Dublin has become an increasingly important location for these kinds of conversations given Ireland’s role as the European base for many of the world’s largest technology companies. Google, Meta, Microsoft and dozens of other major players have significant operations here, which means Ireland sits right at the intersection of tech policy, regulation and innovation in a way few other countries do.
The RDS summit brought together researchers, policymakers and industry figures to work through questions that governments and companies are only beginning to get their heads around. How do you regulate technology that moves faster than legislation can keep up with? Who is responsible when an AI system causes harm? And how do you build systems that are powerful enough to be useful without being so powerful they become unpredictable?
Hassabis has previously spoken about the need for safety to be built into AI development from the ground up rather than bolted on afterwards. His framing of recent hacking incidents as warning shots suggests he believes there is still time to get ahead of the problem, but that the window for doing so is not unlimited.
For Ireland specifically, the summit raises questions about what role the country wants to play as AI becomes more central to how the world operates. Ireland already regulates much of Big Tech’s European activity through the Data Protection Commission, and pressure has been growing for that oversight to extend more meaningfully into AI systems and their potential harms.
Whether policymakers here take the warnings from events like the RDS summit and translate them into concrete action remains to be seen. But with one of the most prominent figures in global AI standing in Dublin and telling the room that what we have seen so far is just the beginning, it is hard to argue the issue can be put off for much longer.