Here is the scariest sentence in cybersecurity right now: "The agentic attackers are coming."

That is not from a sci-fi script. That is Shlomo Kramer, founder and CEO of Cato Networks, reacting to what Anthropic quietly told governments and security firms about its upcoming model, Mythos. The company leaked a draft blog post warning that Mythos can exploit software vulnerabilities "in ways that far outpace the efforts of defenders." Not sometimes. Not in theory. Right now, in testing.

CNN first reported the warnings, and the details are grim. Anthropic is privately briefing government officials about the potential for large-scale cyberattacks enabled by Mythos. The company is also letting certain organizations test the model early to "improve their systems against the impending wave of AI-driven exploits." When the company building the weapon tells the world to start building shields, you pay attention.

For context: Claude Opus 4.6 already discovered 22 Firefox vulnerabilities during a two-week collaboration with Mozilla. Fourteen were rated high severity. Mythos is supposed to be dramatically more capable than that. Now imagine that power in the hands of a criminal group that does not have a responsible disclosure policy.

The Economics of Hacking Just Broke

The real problem is not that Mythos is powerful. It is that Mythos changes the math. A single AI agent can scan for vulnerabilities, test exploits, and keep probing without sleep, without pay, without the fatigue that limits human hackers. Kramer put it bluntly: this is "a watershed event in the history of cybersecurity."

Security researchers have already documented a Russian-speaking cybercriminal who used multiple AI tools to compromise over 600 devices across 55 countries. The attacker had limited technical ability. The AI did the heavy lifting. That was with current models. Mythos and its successors will make that look like a warm-up.

Behind Mythos Is Everything Else

Kramer said something that should keep every CISO up at night: "Behind Mythos is the next OpenAI model, and the next Google Gemini, and a few months behind them are the open-source Chinese models." This is not one company creating one dangerous tool. This is the entire industry on a trajectory where offense permanently outpaces defense.

Defenders have to find every hole. Attackers only need one. When the attacker never sleeps and costs almost nothing to operate, the math is catastrophic. You need to run as fast as you can just to stay in the same place. Anthropic knows this. It built the thing and it is telling the world to brace for impact. That is either admirable transparency or the most terrifying product launch in tech history. Probably both.

First reported by CNN. Additional reporting from Financial World and Penligent AI security analysis.