Anthropic recently agreed to give the European Union (EU) access to Mythos – its highly advanced cybersecurity-focused AI model – after months of discussions between the company and EU officials. Anthropic initially released a limited version of Claude Mythos in April. It reportedly uncovered zero-day vulnerabilities across every major operating system, including one security flaw that had remained hidden for 27 years. (Zero-days are hidden software flaws that hackers can exploit to steal data, take over systems, or launch cyberattacks.) Because of its capabilities, Anthropic immediately restricted access to the model to a small group of organizations involved in critical infrastructure, finance, and cybersecurity under its “Project Glasswing” initiative. So, the EU gaining access to Mythos signals a broader shift in the AI race and where capital may start to flow. Here’s why... Mythos features powerful autonomous and agentic capabilities. Many of you have heard the term “agentic AI” for months, but the model is a concrete example of why it matters. To briefly review, agentic systems are the “next generation” of AI technologies that can make decisions by themselves and adapt to changes. They are essentially like the brain behind a smart assistant or AI robot, able to perceive their environment and act accordingly. Mythos is agentic AI applied to cybersecurity. Traditionally, cybersecurity has relied on highly skilled analysts spending weeks or months searching for vulnerabilities. But if AI can discover software vulnerabilities that humans missed, the opportunity changes from “AI helping workers” to “AI performing work that previously required highly specialized experts.” That shift points to cybersecurity AI as an emerging market in its own right. In other words, the next phase of the AI boom may not be about helping workers write emails faster... but about helping governments and corporations protect digital infrastructure, like data centers and software systems, from increasingly sophisticated threats. And governments getting access to models like Mythos suggests that advanced AI is being viewed less like normal software and more like critical infrastructure, similar to cloud systems, or even energy systems. This could drive more spending on AI security and defense tools that will most likely be run by agentic AI themselves. And if Mythos is any indication, the real opportunity in agentic AI extends even beyond cybersecurity. The same autonomous capabilities that can identify software vulnerabilities could eventually be applied to scientific research, engineering, healthcare, manufacturing, and countless other fields that rely on specialized expertise. That means agentic AI has the potential to become a new kind of digital worker that helps organizations solve complex problems faster, cheaper, and at a much larger scale than before.

Source: InvestorPlace

Read Original Source →