An X user known as 'Impulsive' exposed a massive security failure at the Swedish AI startup Lovable on Monday, claiming they could access other users' code and chat histories. Lovable, which was recently valued at $6.6 billion, denied there was a breach, framing the exposure of public projects as a 'deliberate decision.' The startup is at the center of the 'vibe coding' movement, where non-technical users use AI to build complex software through plain English. However, security experts like Moore warn that these tools heighten risks if users do not understand what data is being exposed. This incident highlights the 'hallucination tax' of automated engineering, where efficiency comes at the cost of security and professional standards. Vibe coding is being marketed as the ultimate democratization of software, yet it often results in the deskilling of the professional class. The Lovable breach exposed not just code, but the fragility of the entire 'agentic AI' moment. As startups like Lovable and Bolt raise billions to replace traditional software engineering, they are simultaneously poisoning the digital commons with vulnerable, AI-generated code. Companies are being urged by investors like Andreessen Horowitz to limit AI use in sensitive business areas, yet the rush toward automation remains unchecked. This is the digital perimeter of cognitive enclosure: a world where software is built by algorithms that neither the user nor the creator fully control. The 'vibe' of innovation is being used to mask the structural risks of handing the infrastructure of the internet over to black-box models. For the workers being replaced by these platforms, the Lovable leak is a reminder that the tools meant to liberate them are often built on shifting, insecure sands.