Anthropic has opened limited access to its new AI model, Mythos, designed for detecting critical vulnerabilities and countering cyber threats. The tool will be available only to select corporations, as its hacking skills are deemed too dangerous for public release.
According to according to According to The Wall Street Journal, approximately 50 companies and organizations that support critical IT infrastructure will receive a preliminary version of the neural network, including Amazon, Apple, Google, and the Linux Foundation. According to Logan Graham, head of the Frontier Red Team at Anthropic, the AI model Mythos has demonstrated such high abilities in finding and exploiting software bugs that the developers currently have no plans to release it to a broader audience.
Cybersecurity experts are increasingly concerned that artificial intelligence is rapidly reducing the window of time between discovering a vulnerability and creating working software to exploit it. At the end of last year, researchers from Stanford University found that neural networks search for and use bugs in real networks almost as effectively as human experts. A notable example was Anthropic's previous model, Claude Opus 4.6, which found more critical vulnerabilities in the Firefox browser in just two weeks than is typically identified worldwide in two months.
The new model Mythos has proven to be about 10 times more efficient than previous AI systems in calculating the cost of detecting a single bug. Although Anthropic does not intend to release its product to the public yet, Logan Graham is confident that similar hacking capabilities will emerge in other neural networks in the coming years. In this regard, the expert urges the IT community to start preparing now for a reality where the delay between discovering a software vulnerability and its application in an attack will be reduced to zero.
Source:
Source: 3dnews.ru
