Tag
Researchers have demonstrated adaptive computer worms powered by open-weight LLMs that can generate target-specific attacks and self-replicate, using stolen compute to evade conventional AI safeguards.