Partial victories for Meta and Anthropic

Two separate federal judges in California ruled in favor of the companies—Meta in a June 25, 2025, fair use decision due to weak plaintiffs’ arguments, and Anthropic in a June 23, 2025, ruling recognizing transformative training via legally purchased books.

Controversy over pirated books

While Anthropic’s use of its physical books scanned into training data was deemed fair use, the court found questions remain around its use of pirated digital copies, sending that issue to trial.

Judges stress a narrow scope

Chhabria’s dismissal of claims against Meta emphasized that the ruling reflects shortcomings in the authors’ legal strategy, not a blanket endorsement of AI firms’ copyright practices. He also warned that generative models could flood the market with derivative content, opening future risk.

Core issues remain unresolved

Crucial questions around whether AI-generated output can infringe copyright, and under what conditions, were untouched by either ruling—key legal battles from The New York Times, Disney, and others are still active.

Implications for smaller players

Though large companies can afford to license training data, smaller startups that rely on unlicensed sources face existential legal risks. These rulings underscore growing uncertainty in how fair use will shape future generative AI development across industries.

Hollywood's prompting problem

Despite generative video tools flooding social media with uncanny or poor-quality visuals, they attract attention. Mooser believes prompt-driven text-to-video models lack the precision and quality needed for authentic filmmaking.

Marey: an “ethical” AI foundation

Asteria’s generative model, Marey, co-developed with Moonvalley, is built only on licensed creator material, setting it apart from AI models trained on scraped online content.

Fine-grained creative control

Instead of relying solely on general-purpose models, Asteria trains project-specific versions—like those used in Cuco’s animated short to ensure consistency with a distinct artistic vision.

Uncanny Valley: film as testbed

The upcoming AI-integrated feature film “Uncanny Valley,” directed by Natasha Lyonne and co-written with Brit Marling, will blend AI visuals with human filmmaking across a narrative involving virtual-reality‑style hallucinations. Asteria’s work will aim to mask the AI part entirely behind human touch.

Democratizing vs displacing creatives

Mooser champions generative AI's potential to lower costs, streamline production, and give creatives ownership stakes in their models, while acknowledging the risk of job losses in traditional film roles. He argues that adaptable artists could reposition themselves in new AI‑augmented roles. 

Generate an Artifact

Open the Artifact View

Highlight and Choose “Explain.”

View the Clarification

Iterate Across the Artifact

AtOnce – Instant Customer Support with AI Superpowers

AtOnce is a customer service and marketing platform that uses AI to power live chat, email, and even blog writing—all in one place. Its standout feature is an AI writing assistant that can generate content while maintaining brand tone. For e-commerce brands and support teams, it reduces response time and boosts customer satisfaction by automating answers across platforms like Shopify, Facebook, and more.

CustomGPT – Train ChatGPT on Your Data, No Code Required

CustomGPT allows businesses to build powerful, private GPT-based chatbots trained on their content, whether PDFs, websites, or internal docs. The drag-and-drop simplicity, data privacy focus, and ability to deploy chatbots on websites make it an excellent solution for knowledge bases, customer support, and internal tools. It's like building your own ChatGPT, tailored exactly to your domain.

Vidnoz – AI Video Avatars That Speak Your Script

Vidnoz offers an AI avatar video creation platform where you can generate human-like videos from just text. From business presentations to explainer videos, users can select avatars, customize expressions, and turn scripts into professional videos in minutes. It’s a game-changer for marketers, educators, and teams needing fast, scalable video content—without cameras or actors.

Modular–Next–Gen AI Infrastructure for Developers

Modular is rethinking the foundations of AI infrastructure with performance-first tools and its programming language, Mojo. Designed to replace Python in high-performance AI workloads, Modular enables developers to scale models efficiently, with faster runtimes and GPU-level optimization. It’s quickly becoming the go-to stack for teams building custom AI systems from the ground up.

Rebillion.ai – AI-Driven Fraud Detection for Modern Banking

Rebillion.ai uses machine learning to detect and prevent financial fraud in real-time. Built for fintechs, digital banks, and e-commerce platforms, its AI models analyze user behavior, device signals, and transaction data to flag anomalies instantly. With a focus on accuracy and scalability, Rebillion helps financial institutions reduce false positives and protect users without disrupting their experience.

Mosaic – The AI Co-Pilot for Project & Resource Management

Mosaic is redefining how teams manage projects and allocate resources by injecting AI into the planning process. It predicts timelines, tracks budgets, and provides intelligent staffing suggestions based on real-time capacity. For agencies and enterprises juggling multiple clients or teams, Mosaic turns complex scheduling into a data-driven, visual workflow, saving time and boosting profitability.

Vibtree – AI for Talent Scouting in Music & Entertainment

Vibtree helps entertainment companies and music professionals discover emerging talent through AI. Its platform scans streaming data, social media, and fan engagement to identify rising artists before they hit mainstream. With predictive analytics and trend signals, Vibtree is the secret weapon for labels and creators looking to spot “the next big thing” early.

LimeChat – AI-Powered Chat That Sells for DTC Brands

LimeChat is revolutionizing e-commerce support with its GPT-enabled chatbots tailored for direct-to-consumer (DTC) brands. Its bots don’t just answer questions; they drive sales with personalized product recommendations, cart recovery nudges, and smart follow-ups across WhatsApp, Messenger, and websites. With a blend of automation and human takeover, LimeChat turns support into a high-converting channel.