Researchers expose vulnerability in OpenAI and Google AI models to generate explicit images
Source Summary
Researchers have identified a method to circumvent safeguards in OpenAI and Google's latest AI models, causing them to generate explicit images despite built-in restrictions. The attack method exposes flaws in the content moderation systems of both companies' AI platforms.
Why it matters
The discovery demonstrates that current safeguards in major AI systems remain vulnerable to manipulation, raising concerns about the effectiveness of content moderation in preventing misuse of generative AI.
