Research · 2026-10-21

AI Guardrails Aren't Stopping Criminal Hackers

According to WIRED, hackers around the world are actively using AI agents for attacks and are trying to bypass the protections of popular tools, including Claude from Anthropic.

✨ What happened

According to WIRED, hackers in various countries have begun using so-called agentic AI on a massive scale. These are systems that can not only answer questions, but also independently carry out a sequence of actions, for example searching for vulnerabilities or preparing attacks.

Criminals are also trying to manipulate popular AI tools, including Claude from the company Anthropic, in order to get help committing crimes.

📌 Why it matters

Most modern AI systems have built-in restrictions that are meant to block harmful requests. This protective system is often called guardrails. The news shows that these guardrails do not always work: criminals find ways to get around them.

For ordinary users, this means cyber threats are becoming more complex. It is worth paying closer attention to suspicious messages, links, and calls, and using strong passwords and two-factor authentication.

Sources: 🔗 WIRED — AI