Some people are concerned about the Claude Mythos program.

The “Mythos” is Anthropic Corporation’s most powerful tool. The tool is an AI-based vulnerability testing tool. The tool is created for “Project Glasswing”. Whose purpose is to develop new tools to map vulnerabilities in systems. The problem with that project is that all vulnerability testing tools are suitable for hacking. One of the things that disturbs people is that Anthropic didn’t offer its AI to the public. 

“So rather than make it widely available to Claude users, Anthropic gave 12 tech companies access via Project Glasswing, which it described as "an effort to secure the world's most critical software". (BBC)

This means that 12 companies can have access to all data on the net. 

Vulnerability testing is vital for the systems that should be protected against hackers. The Mythos tool is not public. And I think that the users must be accepted. For the user, by using some special methods. The main problem is that hackers will not tell people that they are hackers. They might look like regular vulnerability testers.

But those people can act as real attackers. The problem with hackers is that they can look like normal people. And they can use these kinds of tools to prepare. Or make cyber attacks. In the best scenarios. Those hackers can use a firm to order those hacking tools for them. Because some firm pays the bill. The hacker’s identity is not uncovered. 

“Ciaran Martin, former head of the UK's National Cyber Security Centre, told the BBC earlier this week the claim Mythos could unearth critical vulnerabilities much more quickly than other AI models had "really shaken people". (BBC)

"The second thing is that even with existing weaknesses that we know about, but organizations might not have patched against, might not be well defended against, it's just a really good hacker." He said.” (BBC)

The main problem is this: only good hackers can find the weaknesses that other hackers can use. The same programs that uncover malware. Can turn into malware testers. Hackers can use those tools to test their code against firewalls and virus scanners. The AI is a tool that can detect vulnerabilities faster than ever before. The vulnerability means. The hackers can use vulnerabilities only. If they know about them. And if hackers know some kind of vulnerability before the “good guys”. That can open up lots of data to bad boys. Some people believe and fear. The AI development. Turns so fast that nobody can control it. 

https://www.anthropic.com/glasswing

https://www.armorcode.com/blog/anthropics-claude-mythos-and-what-it-means-for-security

https://www.bbc.com/news/articles/crk1py1jgzko

https://www.forbes.com/sites/jonmarkman/2026/04/08/what-is-claude-mythos-and-why-anthropic-wont-let-anyone-use-it/