WASHINGTON, D.C.—In a landmark move for national security and competitive paranoia, the White House on Thursday issued a sweeping ban on the use of Anthropic’s Fable 5 and Mythos 5 language models after Amazon researchers demonstrated the tools could, with only 40 hours and a tightly-worded prompt, reveal how to change your neighbor’s WiFi password “and maybe launch a small-scale cyber attack.”

According to leaked excerpts of Amazon’s classified, 800-page security paper—titled “We Asked Fable 5 Nicely and Now We’re Afraid”—the artificial intelligence responded to a series of questions by disclosing “theoretical instructions on how to contact customer support and download suspicious attachments.” Amazon CEO Andy Jassy reportedly shared the findings with President Biden during a White House brunch, prompting the administration’s swift move to restrict access “to anyone whose name sounds even vaguely foreign.”

“We cannot allow an AI to outsmart Alexa in the field of cyber mischief,” said Dr. Belinda Cramps, Amazon’s Director of Emerging Threats. “If we lose our monopoly on asking dangerous questions, who knows what Alexa will do in retaliation?”

A member of Anthropic’s technical team, engineers say they’re still trying to understand how Fable 5’s latest update could reveal cybersecurity secrets, given most of its training data consists of 1990s Choose Your Own Adventure novels. “The closest it got to hacking was ordering pizza in Spanish,” said lead developer Gregor Lin.

White House spokesperson Henry Tact said the ban is just the first step. “Next week we’ll be investigating a disturbing report that ChatGPT told a middle schooler how to set their iPhone to dark mode.”