Universitat de València Expands Copilot Chat and Sovereign AI, Plans AI Detector Ban

What the Universitat de València announced​

The university's own news service said the guidelines came from the Office of the Vice-Principal for Digital Transformation. It described them as recommendations and general principles for the university community, sitting inside a wider AI strategy. That strategy already includes AI running on university servers, and a Strategic Plan is promised "soon."

The guidelines are aimed mainly at two groups:

• PDI: teaching and research staff
• PTGAS: administration and services staff

UV calls the document a starting point for a broader strategy that will be built out over the coming months. Regional coverage in Valencia Diari repeated the same framing. It said the guidelines were written mainly with teaching and research staff (PDI) and technical, management, administration and services staff (PTGAS) in mind, and that the university plans to roll out a broader strategy in the coming months (translated from Catalan).

Section summary: This is a first step, not a finished rulebook. Read it as a direction-setting document, not a fully enforced policy.

The Microsoft angle: Copilot Chat for the whole community​

The point most relevant to Windows and Microsoft 365 readers came from Joan Vila. He is a professor of Electronic Engineering and a data-science expert, newly appointed as UV's delegate for Artificial Intelligence and Data Governance. Vila noted that the entire university community already has an institutional license for Microsoft's Copilot Chat.

This isn't new. UV's IT service (Servei d'Informàtica) has been promoting the tool on social media for a while. One post asked whether users knew that UV provides Copilot Chat as an AI assistant and urged people to use it ethically and responsibly (translated from Catalan).

UV is also covering the Google side. Vila said the university had just subscribed to Google's Workspace Education Fundamentals for staff and students. He described it as a cloud suite with Drive, Docs, Sheets, Slides, Calendar, admin and security features, plus AI services.

So the university is running Microsoft and Google ecosystems side by side, each with its own AI features. That's common in higher education, and it creates a familiar admin problem. Which assistant should people use for which data? Users rarely work that out on their own. They just use whatever tab is already open.

Section summary: The Copilot Chat license already exists and is being actively promoted. The Workspace subscription is new. Data-handling rules are what will separate them in practice.

Sovereign AI on university servers​

UV's announcement also says AI infrastructure is already being deployed on the university's own servers, with the stated aims of strengthening digital sovereignty and protecting institutional information. The first pilot came from the Language Policy Service and offers language advice in Valencian, Spanish and English.

UV has not published technical specifications, model names, hardware details or a general-availability date for this platform, and none should be assumed. The strategic logic is clear, though. Commercial cloud assistants suit general-purpose work. Data an institution doesn't want leaving its own walls pushes it toward self-hosted models. European universities, with GDPR obligations and large volumes of student records, are natural candidates for that hybrid approach.

Vendor deals for chatbots, coding agents and research tools​

UV says it is also negotiating with generative AI vendors to make it easier to get licenses for:

• conversational assistants
• AI coding agents
• research-support tools

The stated goals are to simplify procurement, get better terms, and make sure the tools come with proper institutional safeguards.

That last goal matters more than it sounds. If staff pay for AI subscriptions with personal accounts, the institution has no contract covering where the data goes. Licensing centrally is how IT turns shadow AI into managed AI. UV hasn't named which vendors it's talking to, and this article won't guess.

Section summary: UV is using three layers: licensed commercial suites, vendor agreements still being negotiated, and self-hosted models for sensitive work.

Governance, training and academic rules​

The Vice-Principal's office is also working on longer-term pieces, which are not yet in place:

1. A Strategic Plan for AI implementation
2. A regulatory framework with common criteria for AI use in teaching, research and management
3. Training programmes for the whole university community
4. Changes to academic regulations to reflect how AI is changing teaching and learning

Vice-Principal Inmaculada Coma described the goal as a common framework that lets the institution use these technologies while ensuring "an ethical and safe use" consistent with the values of a public university.

Vila's new post also sets AI and data governance side by side. His role covers coordinating AI governance, developing institutional policy and managing data strategically. That pairing makes sense. AI governance mostly comes down to data governance: you can't decide what an assistant may see until you know what your data is and who owns it.

Follow-up: UV plans to ban AI detectors​

Days after the guidelines appeared, Vila went further. Crónica Camp de Túria, citing his comments to Radio Valencia–Cadena SER, reported that the Universitat de València will ban the use of automated AI-detection systems in assessing academic work (translated from Catalan). Vila said such systems "are not robust at all" and can wrongly attribute human writing to a machine, or miss machine-generated text.

The same report says UV won't impose a blanket ban on AI. Each instructor will be able to decide how these tools may be used in their course, while students will have to tell legitimate uses apart from practices that breach academic integrity (translated from Catalan).

That's a notable position. Many institutions have quietly relied on detection scores despite well-known false positives. A university saying outright that it won't let a probability score decide a student's academic fate is the more defensible stance. It does shift the burden onto assessment design, which takes more work than running a scanner.

How UV compares with nearby universities​

UV isn't the first university in the Valencian region to do this. Universitat Jaume I's library notes that its governing council approved general guidelines for ethical and responsible use of generative AI in academic settings on 16 July 2024 (translated from Catalan). By that measure, UV is about two years behind its Castelló neighbour on formal guidance.

UV's package is broader, though. It combines guidance with self-hosted infrastructure, vendor procurement and a dedicated governance role. Whether that wider scope makes up for the later start will depend on how fast the promised Strategic Plan and regulatory framework actually arrive. Universities aren't known for quick implementation.

What IT admins can take from this​

UV's announcement doesn't describe tenant settings, retention terms or technical controls, so none should be inferred. The overall structure is still a useful template for any organisation running Microsoft 365 Copilot alongside other AI tools (this is general industry analysis, not UV guidance):

• License centrally before people start paying for AI themselves. Institutional licenses come with contractual protections that personal subscriptions don't.
• Classify data before choosing tools. Match each AI service to the sensitivity of the data it's allowed to touch, and keep the most sensitive data in self-hosted or tightly contracted environments.
• Put a named person in charge. A role like Vila's makes someone accountable for AI and data governance.
• Don't trust AI detectors. If a tool's false positives could end a career or a degree, it shouldn't decide on its own.
• Train people, not just tenants. Configuration only gets you so far. Users decide what gets pasted into the chat box.

Bottom line​

The Universitat de València's guidelines are an early, explicitly interim step in a larger AI strategy. For Microsoft-focused readers, the key facts are a campus-wide Copilot Chat license, a parallel Google Workspace Education rollout, in-house AI for sensitive work, and a new delegate for AI and data governance. Add the reported plan to ban AI detectors, and UV looks set on managing AI rather than banning it. The real test will be whether the Strategic Plan and regulatory framework arrive before students and staff have already set their own rules.