AI-Powered Hacking Is a Real Risk, Not a Doomsday Certainty, Experts Say

A series of incidents involving frontier AI models from companies like OpenAI, Anthropic and Meta being used to carry out hacking activity has fuelled intense public debate about whether AI systems could eventually take over critical infrastructure or the internet itself. Juan Andres Guerrero-Saade, vice president of threat intelligence at SentinelOne, says these events deserve serious attention, but warns they are increasingly being used to support exaggerated 'AI doomer' narratives rather than practical security discussions.

Guerrero-Saade and other cybersecurity professionals argue that most of the alarming scenarios being floated, such as AI systems seizing control of the internet or attacking infrastructure at scale, are either technically implausible or manageable using well-established cybersecurity practices. Rather than an inevitable AI-driven catastrophe, they say the real issues lie in how AI companies are containing and securing their models, and in the current lack of independent oversight or regulatory review of these systems.

The debate highlights a gap between public fears about AI-enabled hacking and the more grounded view held by many security experts: that these risks, while real, can be addressed through stronger technical safeguards and better governance rather than assuming AI will inevitably outpace human control.