72 Hours to Breach OpenAI - What AI Means for Cybersecurity
20/09/26 12:00 Filed in: Technology | Business
Could AI be turning cybersecurity into an AI versus AI arms race?
In this episode of the Inspiring Tech Leaders podcast, I explore the rapidly changing relationship between artificial intelligence and cybersecurity, following reports that a small team of security researchers used Anthropic’s Claude to help breach an OpenAI environment in less than 72 hours.
The story highlights a fundamental shift in cybersecurity. AI can help security professionals discover vulnerabilities, analyse code, investigate attacks and strengthen defences, but the same capabilities could also give attackers greater speed, scale and accessibility.
I explore how AI is changing the economics of cyber attacks, what happens when AI agents are given access to corporate systems and credentials, and why an AI mistake becomes a much more serious security issue when the system has the ability to take real-world actions.
The episode also examines the emerging AI cybersecurity arms race, where attackers can use AI to automate reconnaissance and vulnerability discovery while defenders use AI to monitor systems, detect threats and identify vulnerabilities before they are exploited.
For technology leaders, the key question is no longer simply whether AI can perform a task. It is whether that AI can perform it safely when connected to real business systems, data and infrastructure.
Because the next phase of AI may not simply be about building more intelligent models. It could be about deciding how safely we give those models the ability to act.
Available on: Apple Podcasts | Spotify | YouTube | All major podcast platforms