Keith Porcaro

9/x In my paper, I discuss variants on this basic model: a "blind trust" where a data analysis company is held in trust and the shareholders are beneficiaries (like @Truata1 ); managing a multi-party collaboration; and isolating data from bankruptcy risks. #datatrusts

10/x Of course, data is different from other assets. It’s not as obviously ascertainable; its value comes from inferences and discoveries derived from the data; duties of care around data management are underdefined. #datatrusts

11/x I argue that #datatrusts trusts may not be a good fit for regulating users' relationships with a platform. Data can be freely replicated, and a "user trust" may not be able to prevent users from cutting side deals, undermining the negotiating leverage of the trust.

12/x In modern trusts, (especially in NA), the trust document plays a dominant role in setting the terms of the trust. It can override defaults in trust law. This means that a commercial trust arguably prioritizes a settlor's intent over beneficiary interests. #datatrusts

13/x In a city, for instance, two #datatrusts datatrusts could represent diametrically opposite policy proposals. SidewalkLabs's early proposal is a good example of a data trust being drafted to benefit Sidewalk, rather than restrain them.

14/x Sidewalk’s proposed #datatrusts (which may not be a trust) handcuffs a trustee from putting restrictions on data access, and prevents them from capturing the value of inferences and discoveries made from the data. Sidewalk is best positioned to benefit.

14/x Trusts are a legal tool. It's unfortunate that the name trust has been used to describe arrangements that aren't trusts—it creates unnecessary confusion. But the use of legal tools can help build better norms, in this case around data management. #datatrusts

15/x @thesgconline ‘s use of trusts for reagents, for instance, was aimed at changing norms in science research rather than being a vehicle for legal enforcement. And that's the bigger aim here—building better norms around data management and use. #datatrusts