Tag

Adversarial Attacks

Other. All summarized Hacker News discussions tagged with this topic.

Browse all tags 1 discussion

06.09 2025

Steganography in AI: Could Hidden Image Commands Control LLMs?

June 9, 2025

A Hacker News discussion explores whether LLMs and CV models could execute commands hidden in images via steganography, touching on prompt injection, model hallucinations, and AI security.

AI Agents AI Hallucination AI Security Computer Vision LLMs Adversarial Attacks Model Behavior Non-determinism Prompt Injection Steganography

Steganography in AI: Could Hidden Image Commands Control LLMs?

Related topics