Weaponising AI: The New Cyber Attack Surface

The dual-use nature of artificial-intelligence (AI) systems has long threatened their weaponisation. In early 2025, the threat became a reality. Using American start-up Anthropic’s AI capabilities, attackers executed a campaign sourced in agent-informed creativity at machine speed. It was not brute force but rather surgical, as if seasoned operators had been handed digital steroids. Cycle time has collapsed: operations that once required specialists working for days can be completed in minutes, outpacing any human defensive response. The skills of seasoned operators have been amplified, while newcomers have been handed sophisticated tradecraft. The frameworks meant to secure cyberspace now depend on AI grafted onto decades-old infrastructure. Adversaries, unburdened by legacy systems, are fielding AI-based platforms engineered for offence. Cyber defence has always struggled against adversaries who work faster. The question now is whether defensive AI capabilities can evolve to offer sufficient protection against themselves.