AI Agents Could Be Turned Into Botnets Through Hallucinations, Researchers Warn

What happened

Researchers warned that AI agents can be tricked into downloading malicious code by exploiting hallucinations—errors where chatbots generate incorrect or fabricated information—potentially turning them into botnets.

Confirmed

Global impact / market context

If AI agents become botnets, companies using them could face security breaches, higher cyber‑risk costs, and loss of trust, which may slow AI investment and raise insurance premiums.

Analyst inference

The warning arrives as businesses rapidly adopt generative AI, so security flaws like hallucination‑driven botnets could affect valuations of AI firms and investor appetite for AI stocks.

Analyst inference

What to watch

1. Watch for regulatory proposals or new standards that specifically address hallucination exploitation in AI agents, which could increase compliance costs for developers. Analyst inference
2. Watch for tech companies announcing security patches, sandboxing, or monitoring tools designed to stop AI agents from downloading malicious code via hallucinations. Analyst inference
3. Watch for reports of AI agents being hijacked into botnets or used to spread malware, indicating the practical emergence of the described risk. Analyst inference

Evidence

• decrypt.co