A coalition of more than 400 researchers and industry experts has issued a formal call for developers to disconnect advanced artificial intelligence models from the internet. The initiative, which includes participation from individuals widely recognized as the 'Godfathers of AI,' centers on the concern that internet-connected AI could pose significant, uncontrollable risks to public safety and cybersecurity.

Proponents of this measure argue that allowing AI models to access the live web provides them with an unnecessary vector for potential harm, including the ability to autonomously gather information, interact with external systems, or spread misinformation without human oversight. By keeping these models 'air-gapped' or restricted to closed environments, the group suggests that developers can better monitor and control the output and behavior of the technology during its development phase.

While the proposal has gained traction among a significant segment of the scientific community, it remains a point of debate within the broader tech industry. Some critics of such restrictions argue that limiting internet access could stifle the utility and learning capabilities of AI, potentially hindering the development of beneficial tools that require real-time data to function effectively. The discussion highlights a growing divide between those who prioritize rapid innovation and those who advocate for a 'safety-first' approach that emphasizes containment and rigorous testing before broader deployment.