Former OpenAI board member Helen Toner discusses concerns over Grok AI generating explicit images on X, highlighting the challenges this poses for regulators and the need for more effective oversight as AI capabilities rapidly advance. She emphasizes the importance of transparency, independent audits, and adaptive regulation, while noting the differing approaches and strengths of the U.S., Europe, and China in the global AI landscape.

Helen Toner, a former OpenAI board member, discusses the recent controversy surrounding Grok AI’s generation of sexually explicit images on X (formerly Twitter) at user request. She frames this as a warning sign for governments, highlighting how the rapid development and deployment of AI technologies create complex challenges for policymakers and the public. Toner notes that while the distribution of explicit AI-generated content on a major platform like X is new in scale, the underlying problem is not new—similar tools have existed for over a year, making it difficult for regulators to keep up.

Toner addresses the regulatory response, mentioning that countries like Malaysia and Indonesia have banned such content and are waiting for clearer regulations. In the U.S., she points to the “Take It Down Act,” which will soon require platforms to offer removal options for explicit content, but notes that implementation delays and the ease of re-uploading content limit its effectiveness. She emphasizes that while these steps are necessary, they may not be sufficient to address the broader risks posed by AI-generated explicit material.

The conversation shifts to the responsibilities of AI companies like OpenAI. Toner notes that OpenAI is moving toward making explicit content more accessible, with plans for features like an “adult mode” and a “teenager mode” for ChatGPT. She stresses the need for greater transparency in advanced AI development and advocates for independent third-party audits, rather than direct government control, as a way to ensure companies are acting responsibly in this fast-moving field.

Looking ahead, Toner predicts a paradigm shift in how people interact with AI by 2026, moving beyond large language models to more autonomous AI agents capable of performing real-world tasks. She explains the importance of continuous learning—where AI systems improve over time through feedback—and “world models,” which would give AI a more grounded understanding of the real world, rather than relying solely on text-based training.

Finally, Toner discusses the global AI landscape, comparing China’s approach to that of the U.S. and Europe. She observes that China is rapidly developing its AI ecosystem, often releasing model weights openly to encourage widespread adoption. While China leads in areas like surveillance and is making strides in robotics, Toner notes that the U.S. generally maintains an edge in most AI applications. She concludes that the international race in AI is nuanced, with different countries excelling in different domains and adopting distinct strategies for advancement and collaboration.