There is a growing fear inside cybersecurity circles that AI is moving faster than the systems meant to control it.
And Anthropic’s latest model is becoming the clearest example of that shift.
At the center of the conversation is Mythos, Anthropic’s most advanced AI model, and the warning is coming from someone who has seen the risks up close.
Kemba Walden, who previously led cybersecurity efforts at the national level, did not soften her words. She described Mythos as a wake up call, warning that it can “hack nearly anything” and that current systems are not prepared for what is coming.
That statement alone shifts the conversation.
For years, AI risk has been discussed in abstract terms, bias, misinformation, automation. But this is different. This is about infrastructure, the actual systems that power banks, governments, and critical services.
And Mythos is not just theoretical.
The model has already shown it can identify vulnerabilities across major operating systems and browsers, sometimes finding flaws that have existed for years without detection.
In controlled tests, it has even carried out complex multi step cyberattack simulations, something that used to require highly skilled human teams.
That is where the concern deepens.
Because once an AI can both find and exploit weaknesses, the gap between defense and attack starts to shrink rapidly.
Walden’s argument is not that AI itself is the problem.
It is that the systems around it are not evolving fast enough.
Cybersecurity today still relies heavily on patching known vulnerabilities, reacting after threats are discovered, and updating systems incrementally. But AI like Mythos changes that pace entirely, accelerating both discovery and potential exploitation.
In her view, this creates a dangerous imbalance.
The tools to break systems are becoming faster and more powerful, while the systems to defend them remain slow, fragmented, and often outdated.
That gap becomes even more critical when you look at infrastructure.
Power grids, financial systems, healthcare networks, these are not environments that can be easily updated or rebuilt. Many of them run on legacy systems that were never designed to withstand AI driven attacks.
And that is where the real risk sits.
Not in the existence of AI, but in the mismatch between what AI can do and what institutions are prepared for.
There is also a strategic layer to this.
Anthropic is not releasing Mythos publicly. Access is tightly controlled and limited to select organizations working on cybersecurity. That tells you the company itself understands the potential risk if such a system were widely available.
But even that control is not guaranteed.
Reports have already shown how difficult it is to fully contain powerful AI systems once they exist, raising questions about whether restriction alone is a long term solution.
What this moment signals is a turning point.
AI is no longer just enhancing productivity or creativity. It is becoming a force that can directly impact national security, economic stability, and critical infrastructure.
And that changes the stakes entirely.
So the real question is no longer whether AI will transform cybersecurity.
It is whether the world can build defenses fast enough before these systems outpace our ability to control them.
