Botnets can be created via popular AI tools

Vad har hänt

A study published on July 22, 2026, shows that nine popular AI tools can be exploited to create botnets. The attack is based on large language models (LLMs) generating incorrect but plausible answers that lead users to malicious websites. The method is named "HalluSquatting" and targets the LLMs' inability to indicate that they lack information.

Varför det spelar roll

The security flaw poses a risk of users unknowingly installing malware, which can lead to large-scale botnets. This threat underscores the importance of robust security mechanisms in AI development and the need for user awareness. Consequences can range from data loss to large-scale cyberattacks.

Vem påverkas

Users of AI tools, including individuals and companies, can be directly affected through unintended exposure to malicious code. Developers of AI models and platforms face increased responsibility to implement more secure systems. Security researchers are drawing attention to new attack vectors. Companies using AI in their operations should review their security protocols.

EU-status

Not relevant for EU status.

Mer att veta

Among the nine tools included in the study were ChatGPT, Perplexity AI, Google Gemini, Anthropic Claude, Microsoft Copilot, You.com, Brave Leo, Cohere Coral, and Meta AI. All could be exploited for HalluSquatting under test conditions.