Section 06 of 12 / Investigate
Expert Q&A
Prepare focused conversations with legal, licensing and technical specialists.
17Research pages
10Sources cited
4 November 2026Last updated
The complete section
Read the research.
Full answers, practical checks and the limits of the evidence.
01Which experts would give a balanced view?Use a mix of legal academics, licensing practitioners and technical operators.Read →02What should we send before the conversation?Send a one-page, non-confidential archive description: modalities, subject coverage, dates, update frequency, rights cohorts, proposed use, decision deadline and one disputed assumption.Read →031. Which facts would actually change the fair-use analysis for this archive?Suggested respondent: Samuelson, with Lemley as an alternative.Read →042. How do we determine whether old contributor contracts support the proposed grant?Suggested respondent: Samuelson or the owner's retained IP counsel.Read →053. When should a publisher separate training, retrieval and evaluation into different offers?Suggested respondent: Kaufman, cross-checked with owner counsel.Read →064. What evidence distinguishes paid demand from a growing number of bots?Suggested respondent: Panigrahi, followed by an owner-side customer.Read →075. What does a marketplace or collective route leave for the owner to do?Suggested respondent: Kaufman, with a marketplace operator in a separate session.Read →086. How should we calculate the cost of exclusivity?Suggested respondent: Lemley for competition and legal framing; an owner CFO for modelling.Read →097. How can royalties and platform deductions be independently reconciled?Suggested respondent: Panigrahi and the owner's finance lead.Read →108. What can a reservation protocol enforce in practice?Suggested respondent: Le Meur.Read →119. What should an owner disclose to support a model buyer's transparency duties?Suggested respondent: Kaufman for licensing operations; counsel for applicability.Read →1210. What does model unlearning mean for an enforceable exit plan?Suggested respondent: a qualified ML researcher paired with owner counsel; named technical interviewee still to be selected.Read →13How do we turn the interviews into a defensible decision?Create one evidence row per material answer: claim, speaker, relevant interest, permission to publish, supporting document, date, limitations and effect on the shortlist.Read →
References & limitations
14Frequently asked questions
No.
Read →15Gaps for the next research passNo expert interviews, answers, approvals or quotations have been collected; this document must remain labelled an interview agenda.
Read →16SourcesFollow the references used in this section and inspect their scope.
Read →17Research quality assessmentDetailed decision-oriented agenda grounded in verified expert profiles and primary technical/legal context.
Read →Research quality assessment
Editorial self-assessment, not external certification.
9/10Buyer fit
8/10Depth
8/10Search relevance
8/10Sources
Detailed decision-oriented agenda grounded in verified expert profiles and primary technical/legal context. It deliberately contains no fabricated testimony; completed interviews and permissioned supporting records are the main missing evidence.