The United States has six to twelve months before China gains access to advanced artificial intelligence models or creates its own counterparts capable of carrying out large-scale cyberattacks.
The United States has six to twelve months before China gains access to advanced artificial intelligence models or creates its own counterparts capable of carrying out large-scale cyberattacks.
Recent developments, such as Anthropic’s Claude Mythos and OpenAI’s GPT 5.5-Cyber, demonstrate an unprecedented ability to find vulnerabilities in software, forcing governments and tech giants to rethink their approaches to global security.
This is stated in the material of the POLITICO publication. The journalists note that the development of artificial intelligence is happening much faster than regulators have time to develop appropriate laws to control these technologies.
The new AI models are so effective at detecting vulnerabilities that they are actually being considered potential cyberweapons. For example, the Mythos model has already discovered thousands of previously unknown zero-day vulnerabilities in all major operating systems and browsers. Experts fear that China is trying to copy these technologies through so-called “distillation attacks,” where the results of advanced American models are used to train its own systems.
“This is a hurricane warning,” says Rob Lee, chief cybersecurity officer at the SANS Institute. The models are compressing weeks and years of vulnerability scanning down to seconds, completely changing the mathematics of cybersecurity.
To stay ahead of the curve and allow white hat hackers to patch code holes before attackers find them, developers are restricting access to new features. For example, Anthropic declined China's request to test Mythos, instead granting access to 150 trusted organizations from 15 countries to help strengthen their networks.
The US government is also trying to respond to the threat. US President Donald Trump recently signed an executive order requiring companies to voluntarily submit their powerful AI models for government review 30 days before their public release.
At the same time, Beijing is actively investing in its own developments such as DeepSeek V4 and, according to sources, is deepening cooperation in the field of artificial intelligence and cybersecurity with the Russian government, which is causing serious concern in Washington.
