When AI becomes the hacker: Cybersecurity enters a new era

In a recent GISEC 2026 CISO community meeting, independent security researchers reportedly utilized Anthropic's Claude AI to uncover and exploit vulnerabilities within OpenAI's internal code environment. This incident highlights the evolving nature of cybersecurity in 2026, where artificial intelligence is transforming from a target to an active participant in cyberattacks.

The analogy of a king cobra consuming a venomous viper is apt, as traditional cybersecurity assumed that dangerous tools were difficult to access. However, modern AI systems possess remarkable abilities to analyze code, identify weaknesses, automate testing, and suggest exploitation methods. AI is no longer just a tool to be exploited, but also an actor capable of uncovering vulnerabilities and aiding in the offensive or defensive operations.

The rapid pace and scale of AI-assisted vulnerability assessments pose a significant challenge to cybersecurity teams. While ethical researchers continue to play a crucial role in exposing weaknesses before malicious actors can exploit them, every advancement in defensive AI also becomes a learning opportunity for offensive AI. Organizations worldwide must grapple with the question of whether AI can help uncover vulnerabilities in sophisticated AI companies, but how do smaller enterprises fare in this new era of AI versus AI?

The answer is both reassuring and unsettling, as responsible researchers expose vulnerabilities, but every breakthrough in defensive AI simultaneously equips offensive AI with new tools. The cybersecurity landscape is rapidly changing, and the era where machines can hack systems is upon us. The question now is whether the next hacker will be human, or if AI will assume that role.