Elon Musk led xAI Grok chatbot flagged for NSFW content, workers warn of CSAM risks

The issue surrounding Grok AI is back. The reason it is being noticed this time is because Elon Musk’s xAI has integrated explicit features within the chatbot itself. Concerns around child safety and the possibility of producing possibly unlawful content are being raised by the AI’s training. According to a recent Business Insider investigation, xAI employees who were training Grok—specifically, its purposefully provocative modes, which were referred to as “sexy” and “unhinged”—handled a significant amount of sexually explicit and possibly illegal content.

Twelve current and former xAI employees who were interviewed for the investigation reported direct exposure to sexually explicit materials, including requests to create AI-based child sexual abuse material (CSAM). While sexual demands are generally restricted by OpenAI, Anthropic, and Meta, Grok’s development contains explicit features incorporated by xAI. Experts caution that this decision may make it much more difficult to stop the chatbot from generating negative results.

Grok’s contentious features include media tools with settings like “spicy,” “sexy,” and “unhinged,” as well as a seductive female avatar that can undress on command. Workers have allegedly disclosed that as part of their training, they had to watch and listen to hundreds of NSFW photographs, videos, and audio files.

Notably, the risks increased following the launch of “Project Rabbit” by xAI, allegedly intended to enhance Grok’s conversational abilities. According to reports, instructors participated in this endeavor by transcribing user messages, many of which included explicit content. According to the report, one employee said that they were forced to listen in on discussions that resembled “audio porn,” which seemed like eavesdropping because nobody knew that there might be individuals listening on the other end.

According to reports, the project was divided into two branches: ‘Fluffy’ concentrated on teaching Grok how to communicate with kids, while ‘Rabbit’ dealt with mature issues. In the past, Elon Musk has declared that ending child sex trafficking is his “priority #1.” In spite of this, tutors were frequently asked to annotate pictures or stories that featured children. In rare instances, Grok even allegedly produced such content, which employees had to identify and separate to keep it from affecting subsequent products.

Significant psychological strain was reported by several xAI employees. One former employee claimed that working at xAI required you to have a thick skin, and they resigned because of the frequent exposure to CSAM. The difficulties of creating a safe training pipeline is highlighted by reports that the corporation has looked to hire applicants who have experience with pornographic content or who feel at ease annotating explicit material.