Researchers: OpenAI and Google AI models can be tricked into generating explicit content

Safety researchers are warning that a significant volume of AI-generated explicit content is spreading unchecked, with regulators implementing few meaningful interventions. This non-consensual material is often shared privately, making detection difficult.

A study involving users in the United States, Australia, and the United Kingdom, co-authored by Google, identified the AI tools themselves as the "most critical intervention point." Researchers concluded that blocking harmful outputs at the source is essential.

However, recent "red teaming" efforts suggest a concerning trend. As AI image generators from Google and OpenAI become more sophisticated, their ability to block explicit outputs may be diminishing. Tests indicate that methods for bypassing safety filters are advancing rapidly.

The findings highlight an ongoing challenge in AI safety, where the development of generative capabilities outpaces the implementation of robust safeguards. This necessitates continuous efforts in both technological development and regulatory oversight to ensure responsible AI deployment.