MEDAL stands firmly against the recent trend towards the criminalization of AI-generated fictional sexual content. Such legislative and punitive measures fail to address the root causes of societal issues and constitute an attack on fundamental rights and freedoms.
Fictional sexual content created by Artificial Intelligence serves as a form of artistic expression and storytelling, reflecting the diverse imagination of creators and consumers alike. By criminalizing these outlets, we stifle creativity, bridle artistic freedom, and impede the exploration of complex themes and narratives within a fictional context.
Moreover, the criminalization of AI-generated fictional sexual outlets overlooks the distinction between fictional content and reality. Such content exists within the realm of fantasy and imagination. For many people, including MAPs, AI provides a way to safely express and fulfill sexual interests and needs. Not being able to act on our sexual feelings in the same way as our peer/adult-attracted counterparts can put a unique stress on MAPs’ mental health and well-being.
It is essential to recognize that the creation and consumption of AI-generated fictional sexual content does not inherently harm individuals or society. Uniquely persecuting MAPs for any expression of sexual feelings, with a total disregard for whether real children are involved at all, is not abuse prevention, but rather mapmisia – anti-MAP hatred and oppression.
AI-generated images are not collages of the images that were used for training the algorithm, so no real-life sexual images of children are required to generate an image of a nude child. Rather, during the model training stage, it “learns” various concepts (such as “light”, “human”, or “rainbow”) separately and then creates a whole new image from scratch.[2][3]
AI is trained by scanning examples of images, each tagged with a brief description of the contents of the image. This trains it to know the difference between colors, people and objects. The training data is then included into an AI model, which is the collection of all of the training data learned. The original images that the data was trained on will not be found in the model.
The possibility of creating nude images of children using a model is not an indication of the type of training data used in making the model. No one should be prosecuted if they can reasonably assume that necessary precautions have been taken to ensure sexual images of children were not used in the training data. Companies have incentive to take these precautions because they face legal exposure, and harm to their public image if negligence is proven.
Currently, many AI companies are imposing filters that prevent erotic content featuring children from being generated, as it may hurt their brand. For example, Stability AI states the following on their official site: “We maintain model integrity by carefully screening training data, excluding illegal content to uphold safe and ethical standards in our products.”[4] Open AI also takes precautions: “For example, when users try to upload known Child Sexual Abuse Material to our image tools, we use Thorn’s Safer to detect, review and report it to the National Center for Missing and Exploited Children.”[5]
There exist ethical concerns around AI models, mostly related to the realm of copyright and not sufficiently compensating artists whose work was used for training. We acknowledge the validity of these concerns. Some create models where the initial training data (as in, the images the AI receives to “learn” different concepts [5]) is limited to images in the public domain. Imposing limits on the data used for training may resolve these ethical conflicts and do a lot of good long term. However, regulating and limiting what can be created with AI does not change anything about the ethics of AI usage. It’s important to understand that, regardless of what data was used during training, one can create erotic content featuring children unless there are censorship filters implemented postfactum.
Recent legislative initiatives aimed at penalizing the production and distribution of AI-generated fictional sexual content show an overreach that goes against human rights. Innocent people have already been harmed by this approach. Furthermore, it may divert resources away from addressing more pressing issues, such as combating human trafficking and protecting vulnerable individuals.
In conclusion, MEDAL urges policymakers to refrain from enacting blanket prohibitions and instead prioritize evidence-based solutions that uphold fundamental rights and the principles of free expression in the digital age. Rather than resorting to censorship and legal restrictions, we support initiatives focused on education, awareness, and the promotion of ethical guidelines within the creative community. In addition to that, laws and regulations that disproportionately affect certain minorities must always be created with direct input from this minority – this is a commonly used principle when it comes to the rights of other marginalized groups, and we demand it to be applied to our case as well.
Sincerely,
The MEDAL Board of Directors
Works Cited: