Under manipulations, are some AI models harder to audit?

arXiv:2402.09043v1 Announce Type: new
Abstract: Auditors need robust methods to assess the compliance of web platforms with the law. However, since they hardly ever have access to the algorithm, implementation, or training data used by a platform, the problem is harder than a simple metric estimation. Within the recent framework of manipulation-proof auditing, we study in this paper the feasibility of robust audits in realistic settings, in which models exhibit large capacities. We first prove a constraining result: if a …

abstract ai models algorithm arxiv audit compliance cs.lg data framework implementation law platform platforms robust simple the algorithm training training data type web