Shutterstock debuts creative AI development platform

Shutterstock has introduced a suite of artificial intelligence services designed to accelerate the development of next-generation AI systems.

The offering provides model builders with access to specialized training datasets, creative preference data, and evaluation tools, positioning Shutterstock as an active partner in model construction and assessment rather than solely a data provider.

Announced during SF Tech Week, the new services build on Shutterstock’s expertise in multimodal data licensing. The company aims to deliver high-impact, customized content and insights aligned with the creative and technical needs of model developers.

Daniel Mandell, Senior Vice President of Data Licensing and AI, Shutterstock: "Over the past two years, Shutterstock has significantly outperformed industry expectations, driven by rising demand for its multimodal data corpus. But model builders today need more—they need custom, high-quality, human-calibrated signals across the full training lifecycle. With this new suite of AI services, we're formalizing capabilities that leverage Shutterstock's global creative network, professional-grade production studio, and machine learning expertise to deliver end-to-end value for our partners."

The expanded offering provides support across the full model training process. Shutterstock’s capabilities now include:

Custom content production: Through ten production hubs and more than two million creators across 150 countries, Shutterstock delivers scalable datasets across media types and aesthetics, including advanced formats such as multi-camera and stereoscopic content. Structured datasets include before-and-after pairs, production notes, and metadata for high-value training examples.

Creative preference data: The company harnesses input from professional creatives for rubric-based aesthetic evaluations, enabling more nuanced feedback than traditional crowd-sourced signals. This data supports reinforcement learning, reranking, and evaluation pipelines.

Curation and annotation: Machine learning tools combined with human review provide detailed annotations on composition, lighting, edit history, and other creative traits.

Evaluation and insights: Shutterstock’s pipelines include benchmark prompts, safety and brand-fit audits, and live A/B testing to help developers measure and refine model performance.

Shutterstock’s collaborations with leading AI developers, including NVIDIA, Meta, OpenAI, and Runway, have positioned it as a central contributor to the evolution of generative AI. Its extensive, metadata-rich content library provides diverse, well-annotated assets for training across imagery, 3D, text, and video applications.

Michelle Kwon, Head of Operations and Partnerships, Runway: "Shutterstock has been a key strategic partner for us as we develop industry-leading AI models for creatives. In addition to our data partnership, their expertise and feedback have been valuable to improving the overall performance of our technology."

Mandell: "With these AI services, we’re bringing niche datasets and RLHF-style alignment to diffusion and multimodal models. Our global network of creative professionals and our in-house teams make us uniquely suited to deliver preference data, annotations, and evaluation signals needed to close the gap between model capabilities and human expectations."

By combining its creative community, production infrastructure, and data expertise, Shutterstock is positioning itself at the intersection of creativity and technology. The company’s AI services are designed to help model builders produce systems that are safer, more capable, and better aligned with human and creative intent.

www.shutterstock.com

Follow Marketing Report on LinkedIn