Responsible AI Development Policy

We are committed to the responsible development of our visual generative AI technology

At Black Forest Labs, our mission is to develop frontier AI infrastructure for creators and innovators. With a founding team of pioneers in image and video generation, we are pushing the boundaries of visual generative AI. Our models and services enable developers to build exciting new AI applications, supporting a vast community of enterprises, creators, and researchers worldwide.

Our commitments

• We are committed to creators and builders. We focus on developing visual generative AI technology to ignite human creativity and unlock new economic opportunities for the world’s largest impactful ventures. From photographers to architects, from creative agencies to industrial designers, our technology helps enterprises, developers, and creators build new tools to realize their best work.
• We are committed to open innovation. We openly share model weights to promote greater transparency, support downstream experimentation, and expand global access to creative technology.

• We are committed to responsible AI development before, during, and after release. We apply layers of safeguards to encourage visual generative AI technology's responsible and lawful deployment. From training our models to hosting our services, we care about every step of the process.

Our approach to responsible development

We offer capable, visual, and open generative AI models to enterprises, developers, and creators around the world, and we have a responsibility to help developers and enterprises securely deploy this technology.

• Before release. Before training a model, we carefully filter datasets for unsafe content. We work with trusted partners like the Internet Watch Foundation and utilize our own proprietary technology to identify and remove unsafe content. After pre-training, we evaluate our models for unsafe performance and mitigate these behaviors through various post-training techniques.
• During release. We release our models to developers through a hosted API with clear guidelines for secure deployment and unacceptable use. Our most capable open models are released with licenses prohibiting unlawful misuse, including misuse in violation of privacy and biometric laws.
• After release. When law requires, we moderate unlawful content across our API through filters for prompts, uploads, and outputs. We support our developer and enterprise customers to implement the right level of moderation for their intended applications, subject to our usage policy (https://bfl.ai/legal/usage-policy). We monitor patterns in using our models and services after release, and invite feedback from the community at support@blackforestlabs.ai. We welcome ongoing engagement with authorities, developers, and researchers to share intelligence about emerging risks and respond together with effective mitigations.