I think legal jurisdiction matters in discerning these things. The EU or US both have courts that, despite anyones opinion, are regarded as having robust contract enforcement. If Fireworks is domiciled in either then claiming ZDR and instead training on the data would be an enormous financial footgun. There's reasons why companies on both sides of contracts, even those with little to no US or EU activity, agree to use US or EU courts for enforcement.
Hong Kong I think used to be a popular option as well, before the handover from the UK.
Being terminally 'online' can make people cynical about everything, but have to temper things with reality a bit too.
Another factor in the cynicism is that robust enforcement is very 'pay to play'. I think your point is accurate for a company like Fireworks, largely doing business with peers that could afford litigation if necessary.
Good luck to an individual or small business trying to hold OpenAI or Meta to account if they breach contract over training, or do something crazy like copyright infringement on an industrial scale.
…trust me bro.
It’s obviously easier to believe when they’re not training models.
Eh, anyway this whole thing is just an ad:
> Looking to take Ember-1 one step further, and optimize it for your use case? We are also launching training support for Ember-1, enabling enterprises to build customized, token-efficient models tailored to their needs with their own data. The future of open models is specialized models trained on your specific workload.
Probably, I guess, fancy serverless infrastructure actually makes virtually no difference to hosting really large models that people want to use, and “just” being an inference provider for open weight models turns out to have no moat.
So this is a bit of a pivot to “use our training infrastructure too…!” imo.
Pivot? Sure. Go them. Not what I signed up for though. /shrug