- Published on
Anthropic expands access to Mythos to 150 organizations, including healthcare
- Authors
- Name
- aimode.news
- @aimode_news
Anthropic has given 150 organizations access to its most powerful model Mythos, which is good at detecting vulnerabilities in software. Organizations in healthcare and energy companies, among others, can now join, says Anthropic.
The organizations are in more than fifteen countries, Anthropic says. The intention is to further expand the number of countries where organizations have access to Mythos. Anthropic Project calls giving step-by-step access to Mythos Glasswing. For most partners, going offline will affect at least a hundred million people.
Anthropic wants to release Mythos to the general public, but says it will first work on security to ensure that abuse is not possible and does not have major consequences. By giving organizations access now, these leaks should be able to be closed before cyber criminals can find leaks using such AI models.
Too dangerous
Mythos is still in a preview phase. Mythos is a model within Claude that is mainly intended for security research. The model is somewhat controversial because it can easily detect bugs in code repositories, although a significant part of the controversy lies in the marketing hype that Anthropic creates around Mythos. The company officially unveiled the model in April, but then said it would not make it public for the time being because it would be "too dangerous." Tweakers wrote a background article about it.
The claim that a model is 'too good and too dangerous' is often used by AI companies. That's what they said, for example, when GPT-2 came out. Anthropic is now preparing to release the model, but only for paying business customers. They can then use it to test their own software. For example, Anthropic has already given Mozilla access. This is done under the name Project Glasswing.
Anthropic writes in a blog post that it plans to make Mythos generally available "in the near future." The company first wants to set up 'much stronger safety nets', but the statement that Mythos is too dangerous for a public release seems to have been forgotten.
