Synthesia and Cinder partner to moderate AI video before it renders

Synthesia doesn’t wait for a problematic video to exist before deciding whether to allow it. The London-based AI video company, which enables users to generate avatar-led clips from a script, announced on 4 June that it is expanding its trust and safety infrastructure through a partnership with Cinder, a moderation platform designed specifically for AI-generated content. This move reinforces a strategy Synthesia has followed since its founding in 2017: judge the request, not the finished file.

That ordering is what sets the company apart. For years, online platforms have relied on reactive moderation, hosting content first and flagging it later. Synthesia flips that logic, assessing every script against its policies at the moment of generation, before any frame is rendered by the model.

Cinder integrates into this workflow as what the company calls an in-house agent that performs a second pass on each model decision. It gathers context and escalates to a human reviewer only when a genuine judgment call is required. Every action taken by a reviewer retrains the system, and Cinder’s classifiers support more than 100 languages out of the box.

The rationale for the partnership is sheer volume. In 2025, Synthesia’s automated tools reviewed more than 11.5 million pieces of content and removed 841,957 that violated its policies. Manual review handled an additional 382,792 items, removing 70,272 of them.

Automated review volume grew roughly 77% year over year, from 6.56 million items to 11.58 million. Meanwhile, the number of items reaching human reviewers dropped from 792,586 to 441,086. This pattern is intentional: push clear-cut cases to machines and reserve people for appeals and ambiguity. The firm processed 12,450 user appeals in 2025, reversing about 31% of them upon review.

Synthesia treats safety as a commercial asset rather than a cost. Its customer base skews heavily toward enterprise clients, including more than 60% of the Fortune 100 across financial services, healthcare, defense, and the public sector. For these buyers, a platform that could be turned into a deepfake engine is simply unusable. The company holds ISO 27001, ISO 42001, and ISO 27701 certifications and publishes an annual responsible-creation report it calls Futuresafe.

A recent external test gave Synthesia’s approach a concrete edge. When Belgian public broadcaster RTBF investigated AI video tools for a report titled IA, la fabrique à arnaques (AI, the scam factory), Synthesia blocked the creation of non-consensual deepfakes, political propaganda, racist content, and crypto investment scams. Several rival platforms, by contrast, generated all of them.

For a company that has built its enterprise positioning on refusing certain prompts, an independent broadcaster documenting that refusal is the kind of validation marketing cannot buy.

The Cinder partnership, in Synthesia’s telling, is about keeping that record intact as the product accelerates. As the company ships new avatar types, languages, and surfaces, its trust and safety team wants moderation that moves at the pace of the roadmap rather than lagging behind it.

Whether the combined system holds up under the scale Synthesia is targeting is a question only its future Futuresafe reports will answer. But the bet underneath this announcement is older than the partnership itself: in AI video, the safest moment to say no is before anything is made.