Generative AI, particularly its text-to-image application, has attracted interest from both venture capitalists and lawyers as of late. Two of the most prominent AI art-generating platforms, Midjourney and Stability AI, reportedly infringed on the rights of millions of artists by training their algorithms on web-scraped images, and Getty Images took Stability AI to court for allegedly using images from their site without permission.
This has brought to light the potential ethical and legal pitfalls that generative AI technology can present. However, a startup called Bria is hoping to minimize the risk while still delivering AI-generated visuals with its ethical training model.
Founded in 2020 by Yair Adato and Assa Eldar, Bria was founded to “empower developers and creators while ensuring that our platform is legally and ethically sound”, according to Yair Adato in an interview with TechCrunch. It uses an AI-powered web app and Nvidia’s Picasso cloud to create visuals for social media posts, ads, and other enterprise uses. Bria allows customers to use the “brand guardian” feature to ensure the visuals follow brand guidelines.
Bria’s AI-powered training dataset model is based on licensed images from partners like individual photographers, artists, media companies and stock image repositories. Whenever an image is generated with the platform, Bria looks for the visual in the training set that most contributes to the results to allocate revenue among relevant creators. Additionally, every generated image is watermarked, and the platform provides free access to Non-Profit organizations and academics.
Bria is also planning on releasing an open-source symbol recognition and generative AI art model with in-built attribution features in the upcoming months, which other platforms might use to form similar revenue sharing models with the creators. In addition, Bria has raised $10 million in venture capital and serves an impressive number of companies, including marketing agencies, visual stock repositories, tech and marketing firms.
Yair Adato, who has a PhD in computer science from Ben-Gurion University of the Negev, is confident that Bria’s ethical AI platform is a step up from its competitors. He believes that issues surrounding generative AI can be addressed through using a “high-quality, enterprise-grade, balanced, and safe training set”.
When it comes to content creators wishing to opt-out of Bria’s platform, Adato states that they will be able to. Currently, Bria has its own opt-out mechanism, unlike DeviantArt and Spawning who provide a place for artists to remove their art from popular generative art training datasets.
Bria offers an exciting new solution for businesses seeking to create visuals without facing legal and ethical issues. The startup’s revenue-sharing model and its work on an open-source generative AI model with in-built attribution features might very well lead the way in the industry.
