Hackers can use 9 of the most popular AI tools to assemble massive botnets

In the brief history of AI security, the prompt injection has quickly become the top threat. Large language models are inherently unable to distinguish between legitimate instructions provided by users and malicious ones sneaked into emails, source code, and other third-party content the models are processing. This makes it trivial to surreptitiously inject malicious commands that the LLM readily follows. With no way to enforce this crucial boundary between trusted and untrusted sources, AI engi

Get smart on it

Large language models used in coding assistants have a fundamental flaw: they frequently hallucinate, or fabricate, the incorrect location of code repositories and resources when asked to retrieve them. Attackers can exploit this tendency by registering the fake names that LLMs are most likely to hallucinate and seeding them with malicious code that installs reverse shells on users' devices. This attack, called HalluSquatting, enables criminals to compromise large numbers of devices at scale without targeting each one individually, making it possible to assemble botnets, launch ransomware campaigns, and conduct distributed denial-of-service attacks. The vulnerability stems from LLMs' inability to distinguish between accurate information and fabricated responses, and researchers found the flaw is present across all major LLM models.