Dozens of industry experts have petitioned the White House to restore access to the AI models that currently perform their jobs.
WASHINGTON — A coalition of dozens of top cybersecurity experts petitioned the White House on Tuesday to immediately reverse its export ban on Anthropic’s most powerful AI models, warning that the restrictions have left the nation's digital defenders dangerously reliant on their own skills.
The open letter, signed by senior security architects across the tech industry, argues that President Donald Trump's executive order blocking access to Anthropic's Fable and Mythos models constitutes a severe threat to national security. According to the veterans, depriving the industry of the advanced large language models has catastrophically forced highly paid infrastructure engineers to manually identify vulnerabilities in their own software.
If the administration does not lift these export controls immediately, our entire defense apparatus will be reduced to whatever I can remember from a networking certification I took in 2014.
Prior to the ban, standard industry practice for securing enterprise networks reportedly involved pasting thousands of lines of undocumented legacy code into the Mythos chat interface and typing "make this safe from hackers." Security analysts warned that without the ability to continuously ask the chatbot to patch the very bugs it hallucinated the day before, global supply chains remain uniquely vulnerable to anyone who actually knows how to read an error log.
While the administration argued the export controls were necessary to prevent foreign adversaries from utilizing Anthropic's sophisticated reasoning capabilities to launch cyberattacks, the petitioners noted this logic was fundamentally flawed. The experts stressed that foreign adversaries will not need advanced AI to breach American servers if US defense teams are left staring blankly at command prompts wondering what a firewall rule looks like.
At press time, several prominent cybersecurity firms were reportedly attempting to bypass the ban by asking a lesser AI model if it knew anyone who had a login for Mythos.