US Government stops OpenAI from releasing GPT-5.6 to the public over hacking fears

The American government formally asked OpenAI to keep its upcoming GPT-5.6 model under lock and key, granting access only to a handpicked 'white list' of domestic corporations and federal agencies. According to insider reports, the startup quickly fell in line, shelving its planned public rollout.

This sudden intervention is not an isolated incident, as Anthropic recently had to put its cutting-edge Fable 5 and Mythos 5 models on ice under identical pressure. The official justification centers on preventing potential cyberattacks that these ultra-powerful neural networks could allegedly facilitate.

Cybersecurity professionals have pointed out the obvious flaw in this logic: every single large language model on earth, from Google to the smallest open-source project, suffers from these exact same security blind spots. These flaws are not code bugs that a quick patch can fix, but rather an inherent, unfixable characteristic of how neural networks process information.

The legal basis for this sudden halt traces back to an executive order signed by Trump in early June, which nudged AI developers to voluntarily hand over their models for state inspection 30 days before launch. Since the actual bureaucratic pipeline for these reviews does not exist yet, the administration is simply winging it, blocking major releases one by one as they appear on the horizon.

Treating advanced text generators like weaponized digital plutonium is officially the new normal. Governments are eagerly shifting from reactive policing to preventive bans, content with stopping progress today and figuring out the actual rules sometime tomorrow.