人類組織在偷偷獲取秘密人工智慧網路工具 Mythos 後展開爭奪

Anthropic is a well-known artificial intelligence company that has developed a powerful cyber-security tool called Claude Mythos. This tool is designed to help organizations protect their computer systems from cyber threats. Recently, Anthropic announced that it is investigating a report that a small group of people managed to access the Mythos model without proper authorization. The company explained that the unauthorized access likely happened through a third-party vendor, not through a direct hack of their own systems.

The Mythos model is not available to the general public because of its advanced capabilities. Anthropic has only shared it with select technology and financial companies to help them defend against cyber attacks. However, the company relies on these partners to keep access tightly controlled. According to reports, the people who accessed Mythos had permission to view Anthropic’s AI models because of their work with a contractor, but they used this access in ways they were not supposed to. There is no evidence that these users tried to hack anything or that they used the tool for harmful purposes, but the situation raises concerns about how well advanced AI tools can be protected from misuse.

Cyber-security experts warn that if powerful AI tools like Mythos are used outside their intended controls, they could be used for fraud or other malicious activities. At a recent cyber-security conference in the UK, officials discussed both the risks and benefits of advanced AI. The head of the UK’s National Cyber Security Centre said that AI can make systems safer if used correctly, but it also exposes weaknesses in current security practices. The UK government is encouraging AI companies to work together to protect important networks. Since the most advanced AI models are developed outside the UK, the country depends on companies like Anthropic for access and has little control over how these tools are built or released. The ongoing threat of cyber attacks from other countries makes it even more important to keep these tools secure.