What Happened
Meta has released Muse Spark 1.3 and has not decided whether to publish the model’s weights, The Next Web reports. Meta still plans to release the weights for version 1.2.
The Next Web reports that Meta’s AI chief says the new model moves closer to OpenAI and Anthropic in capability. That is a claim from the vendor rather than an independent benchmark result, so treat it as a positioning statement until third party evaluations appear.
The licensing question sits underneath the meta news here. The EU AI Act exempts genuinely open-source general purpose models from part of Article 53, but The Next Web notes that the exemption does not apply to models classified as carrying systemic risk. In practice, that means a developer cannot simply publish weights and step outside the obligations if a model falls into the systemic risk category.
As general context, model weights are the trained parameters that let you run a model yourself, on your own infrastructure or a provider of your choosing. When weights are published, you gain portability and control. When they are held back, you use the model through the vendor’s API on the vendor’s terms, and your switching costs rise.
Published: Wed, 02 Sep 2026 23:19:09 +0000 Source: The Next Web
What This Means If You Are Buying AI Tools
Most UK marketing teams will never touch model weights directly. You will feel this decision through the tools you buy, because your content platform, chatbot or reporting tool is built on somebody else’s model, and its pricing and availability follow that model’s terms.
So ask your vendors three questions before you sign. Which model or models sit behind the product, can the vendor switch providers without breaking your workflows, and what happens to your data and outputs if that model is withdrawn or repriced. Any supplier that cannot answer clearly is passing their risk to you. When we scope AI marketing services for clients, that dependency map is one of the first things we write down.
There is a UK angle worth noting. The EU AI Act is EU legislation, but plenty of UK businesses sell into EU markets or use tools built by EU-facing vendors, so the obligations reach further than the map suggests. If you operate across both markets, ask your legal adviser where you sit rather than assuming Brexit removed the question.
Key Takeaway
Capability claims move the headlines, but licensing decides what you can actually build on. A slightly weaker model you can run or replace on your own terms is often a safer commercial bet than a stronger one you cannot move away from.
Why London Marketing Company
We treat AI tooling as a procurement decision as much as a creative one. That means documenting which models sit under your stack, where your data goes, and what your exit route looks like if a vendor changes terms, before the tools are embedded in daily work.
We are also realistic about what these tools do for marketing performance. AI speeds up production and analysis. It does not remove the need for a clear offer, sound targeting and enough time in market to read the results properly.
FAQs
Does the EU AI Act apply to my UK business?
It may, if you sell into EU markets or use products from vendors covered by it. The Next Web notes the open-source exemption from part of Article 53 does not apply to models classed as carrying systemic risk, so check your position with a legal adviser.
Why do published model weights matter to a marketing team?
Published weights mean the model can be run outside the vendor’s platform, which affects portability, pricing leverage and continuity if a supplier changes direction.
Should we switch AI tools every time a new model launches?
No. Switching costs time and retraining. Change when your current tool stops meeting a specific need you can name, not because a competitor announced a benchmark.
Talk to London Marketing Company
If you are weighing up AI tools and want an honest read on the dependencies you would be taking on, we are happy to walk through it with you. Talk to our team and we will give you a straight view of what is worth adopting now.
