Here’s what’s on our plate today:

• 🧪 Why Synthesia looks more like Getty Images every day.

• 📰 China's Z.ai targets Mythos, Micron's NVIDIA moment, heatwaves threaten data centers.

• 💬 Prompt of the Day: map the consent risks of licensing AI avatars as stock assets.

• 🗳️ Poll: What's the real question behind Synthesia's pivot?

Let’s dive in. No floaties needed…

The Laboratory

TL;DR

• The Getty mirror: Getty and Shutterstock are merging to defend stock media with AI. Synthesia has spent over a year doing the reverse: licensing Shutterstock's library, taking Adobe Ventures money, and now building moderation infrastructure that only makes sense for a company about to handle far more AI content than its own clients produce.

• Everyone's building a platform: HeyGen now bundles rival models like Sora, Veo, and Kling into its own product. Pika lets outside AI agents tap its video tools directly through Pika MCP: different routes, same destination.

• The stakes: if AI avatars become licensable assets, nobody has settled who controls one once the real person behind it walks away, a gap that gets expensive fast once these marketplaces go live.

Why Synthesia looks more like Getty Images every day

Few industries sit more quietly at the center of modern marketing than stock media. For decades, stock photo, video, music, and illustration companies have served as invisible infrastructure for advertising agencies, publishers, corporations, and small businesses, supplying the visual and creative assets that fill websites, presentations, social media campaigns, training materials, annual reports, and advertisements. The model has worked well for businesses because it allows them to license pre-produced content from vast media libraries, gaining access to millions of images and clips on demand, without commissioning custom photography every time they need a visual asset.

That model also created a multibillion-dollar industry built around ownership, licensing, and distribution. Companies such as Getty Images and Shutterstock became intermediaries between creators and businesses, aggregating content at scale and selling access through individual licenses or subscription plans. Now, generative AI is challenging the assumptions that made those libraries valuable in the first place, enabling marketing teams to generate custom visual assets from text prompts rather than purchasing them. The industry's biggest players are racing to determine what their businesses will look like in a world where media is increasingly generated rather than licensed.

But the more interesting disruption may be running in the opposite direction. Synthesia, a London-based AI company founded in 2017, started by helping enterprises generate videos using AI avatars, and has spent the past 14 months quietly assembling the components of a stock-media platform. Getty and Shutterstock are becoming AI companies. Synthesia appears to be becoming something that looks a great deal like them.

When stock photos learn to talk

Picture the stock photo industry's most enduring character: the generic, smiling professional gesturing at a whiteboard, the same handful of faces recycled across decades of brochures, slide decks, and annual reports. Now picture that person learning to talk, delivering a script in whichever language an organization needs, with gestures that roughly match the words. That is, more or less, what Synthesia has been building: AI avatars, computer-generated presenters that read a script aloud without a camera, a studio, or an actor.

The appeal is easy to understand. Instead of licensing footage of a generic corporate spokesperson, companies can generate one on demand. As of 2025, more than 65k businesses and over 70% of the Fortune 100 were using Synthesia, a figure the company says has since climbed toward 90%. Fortune reported around the same time that AI avatars had become a routine presence inside some of the world's largest companies, underscoring how quickly generated media was moving from novelty to corporate infrastructure.

The speed of that adoption has not gone unnoticed by the incumbents. Getty Images and Shutterstock announced in January 2025 that they were merging, in a deal explicitly designed to fund greater investment in generative AI. At the time, the move looked largely defensive: two incumbents consolidating because the product they had historically sold was becoming increasingly easy to generate from a prompt.

A platform hiding inside a video tool

What is more revealing is what happened next, particularly during the first week of June 2026. On June 4, Synthesia announced a partnership with Cinder, describing it as moderation infrastructure built to handle AI-generated content at scale. A day later, the company introduced an internal governance framework called the 3Rs, intended to impose stricter oversight on how its own AI systems are developed and deployed.

Viewed in isolation, neither announcement seems especially noteworthy. Growing technology companies routinely invest in trust-and-safety tools and governance frameworks as they expand. Taken together, however, they raise a more interesting question: why would a company best known for helping large enterprises create internal training videos suddenly require moderation infrastructure designed for AI-generated content at scale?

The answer matters because it suggests something larger may be happening inside Synthesia. The company is preparing for a future in which it is not merely generating videos for corporate training departments, but operating a platform that creates, manages, and distributes AI-generated people, voices, and media assets at scale. Seen through that lens, the moderation and governance announcements begin to look less like routine housekeeping and more like signs of a company preparing for a much bigger role in the media ecosystem.

The explanation comes into focus once you line up everything else Synthesia has done over roughly the past 14 months. In April 2025, it licensed Shutterstock's content library to help train its newest avatar model. The same month, it took a strategic investment from Adobe Ventures, the venture arm of the company that owns Adobe Stock. In July 2025, it shipped a tool that lets enterprise customers generate their own video and image B-roll, the supplementary footage used to fill out a scene, using Google's Veo 3 model, directly within Synthesia rather than purchasing stock footage elsewhere. And back in March 2025, it had already launched a talent program giving the professional actors behind its stock avatars a feedback channel and early access to new features.

None of this happened in a vacuum. In January 2026, Synthesia raised funding at a $4B valuation, roughly four times its valuation less than three years earlier, when a $90M round valued it near $1B. Individually, each of these moves reads as a partnership, an investment, or a feature update. Together, they describe a company quietly assembling the components of a media business: content to train on, capital from a stock-media incumbent, tools that substitute for stock footage, and a working relationship with the performers whose likenesses populate its catalog.

And while Synthesia was busy laying the groundwork for a possible conversion to a platform, Getty and Shutterstock, which had spent decades operating as content licensing platforms, are now investing heavily in AI to protect and expand their businesses. Yet though they may be moving in different directions, in both cases the end goal is the same: to create a marketplace of reusable assets that customers can call on demand.

The platform race

Synthesia's evolution is not an isolated case. Across the generative-media industry, the largest players are beginning to move beyond selling a single AI capability and toward becoming platforms that sit between creators, businesses, and other AI systems.

HeyGen, Synthesia's most direct competitor in avatar video, has spent the past year folding other companies' generation models directly into its product. Its site now advertises access to OpenAI's Sora, Google's Veo, Kling, and ElevenLabs' voice models, all from inside the HeyGen interface. Rather than building a marketplace for outside creators, HeyGen is becoming an aggregator of other companies' generative tools.

Pika, meanwhile, now describes a wider product universe spanning a video tool, a consumer app that turns a single photo into a short animated effect, and something called Pika MCP, which lets other AI systems, including Claude and OpenAI's models, call on Pika's video generation as a building block inside their own workflows. Pika is positioning itself as infrastructure that other AI agents use, rather than as a tool a person opens directly.

Three companies, three different routes: a creator-facing marketplace, a model aggregator, and an agent-accessible utility. Yet all are converging on the same underlying shift. The AI video tool is becoming a platform that other parties, whether human users, businesses, or other AI systems, plug into.

Who owns the digital person?

Building a platform is not simply a technical challenge. It is also a legal and governance challenge. The moment AI-generated people, voices, and likenesses become assets that can be reused, licensed, distributed, or incorporated into other systems, questions emerge that did not exist when these products were merely standalone software tools.

In the United States, legislation introduced in Congress would create a federal right governing how a person's voice and likeness can be used in AI-generated digital replicas. In the European Union, regulators have begun work on rules requiring AI-generated content to carry labels indicating its origin. And Synthesia's own governance materials, published in April 2026, acknowledge a gap directly: its framework for managing avatar likenesses raises, without fully resolving, the question of what happens to an avatar once the person it was built from leaves the organization that licensed it.

That question gets harder once an avatar becomes something a marketplace can sell to a buyer with no relationship to the original company or the original person. A stock photograph, once licensed, is a frozen artifact: the model in it does not need to do anything further for the license to stay valid. An AI avatar is closer to an ongoing performance, one that can be re-prompted indefinitely by parties whose source it may never meet. If that becomes a normal commercial object, licensed the way a background image or a font is licensed, the underlying question stops being about AI video at all. It becomes a question about what kind of property a face is, and whether the logic of stock media, an asset licensed repeatedly to strangers for purposes its subject never specifically agreed to, was ever built for something that can talk back.

A different kind of library

Right now, the only visible evidence is two days of blog posts about moderation tooling and an internal governance update, sitting alongside a string of partnerships that look unremarkable read separately. Taken together, however, they point to a broader possibility: that generative AI is not simply changing how stock media is created, but also who gets to be a stock-media company. Getty and Shutterstock are racing to adapt their libraries to an AI world. Synthesia is building a different kind of library altogether, one stocked not with photographs and video clips, but with reusable digital people.

The legal and regulatory infrastructure for that library does not yet exist. What it will look like and who will be responsible for it are the more consequential questions behind the product announcements.

Bite-Sized Brains

• China's Z.ai targets Mythos: China's Z.ai released GLM-5.2 with a focus on cybersecurity capabilities, positioning it as a direct challenger to Anthropic's Mythos-class models.

• Micron's NVIDIA moment: Wall Street increasingly sees US memory maker Micron as the next NVIDIA, betting that AI's appetite for high-bandwidth memory makes it a core infrastructure winner.

• Heatwaves threaten data centers: A new analysis warns that intensifying heatwaves are emerging as a serious climate risk to AI data centers, straining cooling systems exactly when demand peaks.

Prompt Of The Day

Tuesday Poll

Rate This Edition