Black Forest Labs' CEO urges G7 leaders to keep AI development open

Robin Rombach, co-founder and CEO of image-and-video model maker Black Forest Labs, used a G7 appearance to argue that governments should actively protect open-weight AI development rather than let safety fears push the industry toward a handful of closed, proprietary labs.

What's new

In a June 18, 2026 post on the company's blog, Black Forest Labs said Rombach spoke alongside world leaders to make the case that open and closed AI are not opposing choices: "Innovation in visual AI can be both open and responsible, not one or the other." The company framed open weights as a matter of national and economic resilience, not just developer convenience: "Open technology is vital for transparency, competition, and strategic independence in AI."

Rombach's central warning was about concentration risk. Overcorrecting toward restriction, he argued, carries its own danger: "A climate of fear around open technology will leave the world reliant on a handful of firms for critical infrastructure." Rather than arguing against regulation outright, Rombach called for governments and industry to make open, responsible development "the norm, not the exception," backed by concrete tools — safety evaluation standards and targeted rules — rather than blanket restriction. He specifically pointed to the US, EU, and UK as needing to lead on combating sexual deepfakes as one area where targeted regulation is warranted, while cautioning against rules broad enough to suppress open model releases generally.

Context

Black Forest Labs, the company behind the FLUX family of open-weight image models, has positioned itself as one of the few frontier-capable labs still shipping open weights for visual generation, in contrast to closed proprietary systems from OpenAI, Google, and Midjourney. That makes the company a direct stakeholder in how G7 governments regulate AI: rules aimed at curbing misuse of image and video generation — particularly deepfakes — land differently on an open-weight provider, whose models can be run and modified outside any central point of control, than on a closed API provider.

The G7 has increasingly become a venue for AI governance coordination, following prior commitments on AI safety codes of conduct and generative-AI risk frameworks. Open-weight advocacy at that level reflects a broader lobbying push by open-model developers — including Meta, Mistral, and others — arguing that export controls and safety rules aimed at frontier labs risk entrenching a small number of closed incumbents.

Why it matters

The deepfake carve-out Rombach endorses is notable coming from an image-model maker: it concedes that some targeted restriction is warranted specifically in the domain his own company's models touch most directly, rather than opposing regulation wholesale. That distinction — targeted rules on harmful use cases versus blanket restriction on open releases — is likely to be the actual fault line in G7 AI policy debates over the next year, more than a simple open-versus-closed argument. Whether governments accept Rombach's framing will shape whether smaller labs can keep releasing frontier-capable open weights at all as safety rules mature.