A framework that finally puts AI on the nondelegable list
Hospital-radiology joint ventures are forming faster than the contracts governing them have kept up, according to a governance framework from Mayur K. Virarkar, MD, MBA, and Dheeraj Reddy Gopireddy, MD, MBA, published in Academic Radiology in 2026 (DOI: 10.1016/j.acra.2026.08.035) and covered by Radiology Business. Its argument, built on principal-agent and stakeholder governance theory, three decades of joint-venture quality literature, and Donabedian's quality framework: when a deal doesn't structurally separate financial incentives from clinical authority, diagnostic quality degrades.
The authors' fix is a 13-domain responsibility matrix that splits authority cleanly: the radiology group keeps nondelegable clinical-quality authority over a defined set of domains, the operating partner gets full authority over day-to-day operations, and a co-chaired Joint Quality Committee sits at the interface between the two. Four of the named nondelegable domains are accreditation, peer review, radiation safety — and, notably for any group running AI-assisted reporting, artificial intelligence approval.
Why "AI approval" being on that short list is the real signal
Putting AI approval next to accreditation and radiation safety is a meaningful choice. It tells a JV partner, in writing, that whether and how AI participates in a diagnostic read is a clinical decision the radiology group does not have to negotiate away for operating capital or scale. That's real protection, and it's more than most JV contracts have ever said about AI explicitly.
But approval authority and infrastructure control are not the same clause. The framework's five contractually enforceable provisions — a volume-quality firewall preventing physician pay or scheduling from being tied to imaging volume, physics veto authority over equipment release, a named arbiter for clinical-versus-business disputes, seven preset escalation thresholds benchmarked to ACR standards (a 3% image-rejection-rate trigger, for example), and audit and termination mechanisms — are all built to protect quality decisions. None of them, as covered, separately addresses who owns the PACS/RIS contract, the reporting-software licensing, or the integration budget that any AI tool has to run on. That's the piece of "operational authority" a JV partner inherits by default.
Where the gap actually bites
A radiology group can hold nondelegable authority to clinically approve an AI reporting tool and still lose the practical fight over which tool it gets to approve. If the operating partner — a hospital system standardizing IT across its network, or a private-equity platform consolidating a portfolio of imaging centers onto one enterprise contract — already licenses a reporting stack across its other sites, "operational authority" routinely comes to mean the vendor decision too, not just staffing and scheduling. AI approval sitting on the clinical side of the ledger doesn't stop that unless the contract separately says the group can select, evaluate, or replace the AI vendor its clinical workflow depends on.
That distinction matters most for groups mid-consolidation who are actively evaluating AI reporting vendors — exactly the position where a contract's silence on technology-stack control turns into a default answer, decided by whoever signs the enterprise IT agreement rather than by clinical fit.
| Contract area | Who the new framework assigns it to | What's typically left unwritten |
|---|---|---|
| Ownership & governance | Negotiated directly (equity split, board seats) | — |
| AI clinical approval | Radiology group (nondelegable) | — |
| Accreditation, peer review, radiation safety | Radiology group (nondelegable) | — |
| Day-to-day operations | JV partner (full authority) | — |
| Reporting/AI technology stack & vendor contract | Not separately named | Falls to whoever holds "operational authority" by default |
| Vendor exit & data portability | Not addressed in the framework as covered | Terms if the venture or vendor relationship ends |
Middle column reflects the domains and provisions described in the 2026 Academic Radiology framework as covered by Radiology Business. Right column is the gap this article argues due diligence should close before signing.
A narrower ask than a full due-diligence checklist
This isn't a call to renegotiate an entire joint venture around technology. It's one clause: a carve-out stating that the radiology group's right to approve AI clinically includes the right to choose, evaluate, and if necessary replace the AI reporting vendor supporting that workflow — separate from, and not overridden by, the partner's general operational authority over IT and procurement. Groups already weighing whether to stay independent rather than sell to private equity face a version of this same question earlier; for those that do enter a joint venture, the technology-stack clause is the mechanism that keeps that choice from being made for them later, quietly, at renewal.
It's also a different risk than the staffing and turnaround-time gaps that have already shown up in radiology consolidation deals: a group can hit every staffing and turnaround benchmark in a contract and still find its AI vendor decision was never really its own, because nobody wrote down that it was supposed to be.
Where xAID fits
Any AI reporting vendor a radiology group approves under a clause like this should be evaluated the same way the group would evaluate one on its own — see xAID's AI vendor evaluation checklist for the criteria that matter regardless of who owns the imaging center. xAID's reports are ready-to-sign, with an in-house radiologist reviewing every preliminary before the client's reading radiologist signs the final, and the model is built to integrate alongside a group's existing reporting workflow rather than require a system-wide platform swap — which is exactly the kind of low-lock-in fit a group should be negotiating the contractual right to choose in the first place.
Frequently asked questions
Does a radiology joint venture contract usually say who controls AI oversight?
A governance framework published in Academic Radiology in 2026 says it should. It assigns artificial intelligence approval to a short list of nondelegable clinical-quality domains — alongside accreditation, peer review, and radiation safety — that must stay with the radiology group no matter how much operational authority a hospital or private-equity partner takes on in the joint venture.
Does AI approval authority protect a radiology group from AI-vendor lock-in in a joint venture?
Only partly. The framework's nondelegable domains and five enforceable provisions (a volume-quality firewall, physics veto authority, a named arbiter, preset escalation thresholds, and audit and termination rights) protect the group's clinical say over which AI tool it approves. None of those provisions, as covered, separately names who controls the underlying reporting technology stack — the PACS/RIS contract, vendor licensing, and integration budget that AI approval authority depends on to mean anything in practice.
What technology-related terms should a radiology group negotiate in a joint venture contract?
Beyond clinical AI approval rights, groups entering a joint venture should negotiate an explicit carve-out for reporting-technology decisions: the right to select or replace the AI reporting vendor supporting its clinical workflow, defined data- and report-portability terms if the venture ends, and protection from being defaulted onto a partner's enterprise-wide platform choice purely for procurement convenience.
Why does reporting-technology control matter for radiology groups evaluating AI vendors during consolidation?
Once a hospital system or private-equity-backed partner takes over day-to-day operations, it typically also inherits procurement authority — including which reporting and AI software the imaging center runs. A group can retain the clinical right to approve an AI tool and still end up standardized onto whatever platform the partner already licenses across its other sites, unless the contract explicitly separates technology-stack control from general operational authority.
Source: Mayur K. Virarkar, MD, MBA, and Dheeraj Reddy Gopireddy, MD, MBA, "A Governance Framework for Diagnostic Quality in Hospital-Radiology Joint Ventures," Academic Radiology (2026), DOI: 10.1016/j.acra.2026.08.035, as covered by Radiology Business. The technology-stack control gap discussed in this article is xAID's analysis of what the covered framework does and does not name, not a claim made directly by the study's authors. Figures are rounded as reported.