Key Takeaways
- Safety researchers warn of widespread AI-generated explicit content.
- Google and OpenAI models increasingly produce unwanted explicit outputs.
- Regulators have not effectively addressed the issue, study finds.
Safety researchers have raised concerns about the proliferation of AI-generated explicit content, which continues to spread unchecked despite efforts by major tech companies to address the issue.
A recent study co-authored by Google researchers found that a significant volume of AI-generated explicit content is being shared through text and direct messages or stored in private clouds, making it difficult for regulators to detect and remove.
The study, which surveyed internet users in the United States, Australia, and the United Kingdom, highlighted the need for more effective interventions at the source—namely, the AI tools used to generate such content.
Google researchers emphasized that the 'most critical intervention point' lies in the AI tools themselves, suggesting that as these tools become more advanced, they may be less effective at preventing the creation of explicit outputs.
Red-teaming exercises conducted in recent months have shown that leading AI image generators from Google and OpenAI are struggling to block explicit content, according to the study.
The researchers noted that while some improvements have been made, the overall trend is concerning, as the tools are becoming more capable but less reliable in filtering out unwanted content.
Regulators have not yet taken significant steps to address the issue, leaving the burden on tech companies to find solutions.
The study's findings underscore the ongoing challenge of balancing innovation in AI with the need to protect users from harmful content.
