A rather interesting piece of news: The Bureau of Industry and Security (BIS) of the U.S. Department of Commerce has officially banned providing access to these models to any foreign nationals worldwide — including even foreign employees of Anthropic itself. The company had only 90 minutes to comply. Since filtering the nationality of hundreds of millions of users in an hour and a half is unrealistic, Anthropic simply shut down both models for everyone.
At first, there was talk of fear of China copying the algorithms, then a political motive was sought in the confrontation with the Pentagon over a government contract — allegedly a petty retaliation for refusing to cooperate. But later the true reasons came to light.
The Economist published an insider account from a closed briefing. Vice Chairman of the Senate Intelligence Committee Mark Warner revealed the assessment of the Mythos model's capabilities provided by the head of the NSA and Cyber Command, General Joshua Rudd:
"[This tool] broke into almost all of our classified systems. And we're not talking about weeks, but a matter of hours."
As it turned out, intelligence agencies conducted a controlled stress test. The AI demonstrated remarkable autonomy: without human assistance, it found hidden vulnerabilities, such as a 17-year-old bug in FreeBSD, and chained them together for an attack.
Many consider this either an overreaction by ministry officials or a controlled leak to justify harsh measures. But the fact remains...
📲 MAX 👮♂️ VK ☺️DZEN 🙃 SETKA ✈️ TG #ai #neuralnetworks #ai #neuralnet #news #trends
Comments
0No comments yet.
Sign in to join the discussion.