alt Hacker NewsDude, your closed source model could be backdoored way more easily simply because it's only available via API and knows whom it is talking to: Just serve your targeted user a different adversarial model.
Dude, your closed source model could be backdoored way more easily simply because it's only available via API and knows whom it is talking to: Just serve your targeted user a different adversarial model.
Both propriety and open weight models are closed source. Open source would include training data so anyone could train a similarly capable model.