By standardizing how & what information is disclosed across the AI supply chain, we provide protection to both upstream and mid/downstream developers. We unpack what this might look like (ideally), along with various other considerations Can't wait to hear what people think! 5/5

Characteristics of AI make identifying liable parties hard. Instead, AI policy has emphasized risk management. Without intervention, that means responsibility will often fall on the “last mile”, not upstream providers. Preventing 1-sided terms of service can help. 4/5

By e.g. subsidizing data marketplaces (w/strict terms of use) & small businesses' compute, we foster competition—providing users with options and diversifying the value systems perpetuated along the supply chains 3/5

There are many issues that are exacerbated when we deploy models through the AI supply chain. Liability, transparency, and optionality are just a few of these. 2/5