Experts Warn: Open-Source LLMs Dramatically Lower the Barrier for Cyber Attacks

Tim_Dettmers · x · 2026-08-01

AI security experts warn that open-source tools are drastically lowering the barrier to entry for cyberattacks. What previously took hackers years to build can now be initiated at scale with a single prompt.

Monitoring gaps: Even without their own GPUs, attackers can exploit mainstream model routers/providers that currently lack large-scale monitoring to stop malicious use.
Asymmetry: There is a fundamental asymmetry in hacking—attackers can fail a million times and win once, while defenders must defend against all attacks 100% of the time.
Capabilities: Although models like Kimi K3 still lag 6-9 months behind US frontier models in cyber capabilities, the cat-and-mouse game will continue as capabilities advance.

Original post →

Experts Warn: Open-Source LLMs Dramatically Lower the Barrier for Cyber Attacks

More from Safety