Why third-party AI flaw disclosure needs real submission infrastructure
A research proposal argues for standard reports, published testing rules, safe harbours and coordination across providers; it is not itself a disclosure route.
“In-House Evaluation Is Not Enough” proposes a framework for third-party AI flaw disclosure. The paper calls for standardised reports, published rules of engagement, broader disclosure programmes and coordination where a flaw transfers across providers.
The paper is background material. It cannot receive a disclosure. Use a currently open reporting programme such as FLARE-AI or an applicable provider channel, and check the programme’s scope and safe-harbour terms before testing.