AI Chiefs Warn on Security Risks

“Recent AI-driven cyberattacks were an early warning shot,” Google DeepMind co-founder Demis Hassabis cautioned. He stressed that more serious incidents could emerge over time unless corporations and governments intensify efforts to secure artificial intelligence systems.

Addressing the International AI Summit in Dublin, Hassabis stated that the industry must adopt safety protocols similar to those utilized in the biological and nuclear sectors. He advocated for measures such as secure, isolated "sandboxes" for testing AI agents before internet connectivity, alongside enhanced cybersecurity frameworks.

“We’ve got to kind of asymmetrically power up our defensive side to keep ahead of the attackers,” he remarked, suggesting that financial institutions and governments could implement AI to identify and remediate software vulnerabilities before systems are released to the public. He also reiterated calls for an international standards organization to oversee AI safety, though he clarified that any such framework must remain flexible enough to accommodate technological progress without hindering innovation.

Microsoft President and CEO Brad Smith also supported the implementation of more robust safeguards, characterizing 2026 as a "sobering year" for AI safety. “No company or technology should be above the law,” he stated, noting that regulation should be practical and balanced while still facilitating innovation. Smith added that firms should be prepared to moderate development when necessary, citing Microsoft’s decision two and a half years ago to reassign 34,000 engineers from product development to cybersecurity. “It was painful, but it was the right thing to do,” he concluded.