Spark's trust model comes with some disadvantages, though.
Rogue employer could change the code to steal funds. Moreover, it is more susceptible to government interventions.
Discussion
Also, it could tend to centralization (imagine LightSpark being the only operator) and lacks privacy, since all tx data is public.
Possible improvements:
- More Spark Entities with better privacy policies
- BOLT12 and BIP353 support
- Improve SSP scaling
- Improve engagement with open-source dev community
- Public proof of no forking happening