Two developments from 23–24 July that are directly relevant to Dutch professionals: Black Forest Labs yesterday launched FLUX 3, a multimodal model for image, video and audio in a single architecture, and the Autoriteit Persoonsgegevens earlier this month published practical GDPR guidance for developers and users of generative AI.
FLUX 3: image, video and audio from one model
On 23 July 2026, Black Forest Labs launched FLUX 3, their first model that combines images, video and audio within a single architecture, unlike earlier systems that placed separate models behind a shared interface.
The core capabilities:
- Video up to 20 seconds with natively synchronised audio, including dialogue and sound effects
- Image generation and editing in multiple styles and resolutions
- FLUX Mimic: an extension for robotic action prediction, developed with Mimic Robotics; Audi is already testing it for production line tasks, where 30 minutes of training data per task is sufficient
In preliminary evaluations, FLUX 3 Video was preferred over Runway Gen-4.5 (77%) and Luma Ray 3.2 (93%). Black Forest Labs itself stresses that "the model and evaluation system are still under development" and that full benchmarks will follow later.
Availability: Video and Action are now available via the API and selected partners (early access). Image early access follows in the coming weeks. An open-weight Dev version is planned for later in 2026.
For content creators, marketers and media professionals, FLUX 3 is not yet self-service: the route in is a partner relationship with Black Forest Labs or access through an API integration. If you want to start now, you can use the more stable FLUX.2 for image work.
See also the model comparator for an overview of image generation models and their prices.
AP: practical GDPR guidance for generative AI
On 14 July 2026, the Dutch Data Protection Authority (Autoriteit Persoonsgegevens) published two long-awaited documents: guidance for developers of generative AI models and a practical checklist for organisations wanting to deploy generative AI.
For developers, the guidance covers when processing personal data is lawful, which legal bases apply, and which safeguards apply to data collection, cleaning, enrichment and use for model training.
For deploying organisations, the checklist offers a step-by-step assessment: should I deploy generative AI? Can I do so responsibly from a technical standpoint? And is it legally permitted?
AP board member Katja Mur summed it up: "Innovation is welcome, but with strong safeguards." The AP explicitly advises deploying generative AI wherever possible without processing personal data.
What this means in practice:
- Organisations using a language model as a work tool (summarising, drafting, analysing) need to be clear about which personal data may end up in prompts
- Developers who scrape or purchase training data must document the legal basis for each category of personal data
- For high-risk systems, the EU AI Act also applies (more on the deadlines on the safety page)
What to do this week?
- FLUX 3 on your radar? Sign up for the waitlist or work via an existing partner platform. Hold off on production use until Image is generally available and pricing is transparent.
- AP GDPR guidance: Download the checklist from the AP website and work through the steps for your main generative AI applications. This is not a one-off task: personal data can creep into prompts in places you would not immediately expect.
- Sector pages: On the sector pages you can see which tools offer EU hosting or ISO 27001 certification, which is relevant to the AP checklist step "is it legally permitted?"
