ChatGPT instead of an expert? The Public Procurement Tribunal says: That's your risk.
A landmark 2025 ruling by the Lower Saxony Procurement Tribunal confirmed that companies bear responsibility when relying on generic AI to review tender documents. The takeaway: combine domain-specific AI with expert knowledge.


Felicitas von Rauch
Founding Growth Lead
A recent ruling shows what happens when companies blindly rely on AI in public tenders. And why this is not an argument against AI, but rather for using it correctly.
AI is fundamentally changing how companies participate in public tenders. Analyzing tender documents, extracting requirements, drafting concepts. What used to take days now takes hours. And trust in the technology is growing: more and more bid teams are using AI, for example as a writing aid. But how far can that trust go? A 2025 ruling by the Lower Saxony Public Procurement Tribunal shows for the first time where the line is.
The case
A company applies for an EU wide tender for chat and voice bots with AI capabilities. The contract volume is large, the procedure complex. A negotiated procedure with a participation competition for a framework agreement. Exactly the kind of procurement process where you should review the documents with particular care.
The company decides to have the tender documents reviewed not by an expert, but exclusively by ChatGPT 4.5. The AI finds nothing objectionable. The company submits its bid.
Then the disappointment: the bid is not accepted. Only now, after the submission deadline has passed and after receiving the rejection letter, does the company notice serious errors in the tender documents. Among other things, a binding maximum quantity was missing from the framework agreement. A point that, according to European Court of Justice case law, must be specified.
The company raises a formal objection and initiates a review procedure before the Lower Saxony Procurement Tribunal.
Objection raised too late, opportunity lost
The Procurement Tribunal rejects the application on the key points as inadmissible. The reasoning: anyone who can identify errors in the tender documents must raise them with the contracting authority before the submission deadline. Failure to do so forfeits the right to complain. This is known as "Rügepräklusion" (objection preclusion). This is not a new rule but established practice in procurement law.
The missing maximum quantity was, in the Tribunal's assessment, objectively recognizable to a company experienced in procurement law. The question of whether a contracting authority must specify a maximum quantity in a framework agreement has been debated for over 25 years. Since a 2021 ECJ ruling at the latest, the answer is clear: yes, they must.
The key passage: ChatGPT is not a legal advisor
Particularly noteworthy is what the Procurement Tribunal says about the role of ChatGPT. The company had argued that the grounds for objection only became apparent after receiving legal counsel. The Tribunal does not accept this and states clearly:
"If a company has the tender documents reviewed merely by a large language model instead of seeking legal advice from a licensed service provider, errors made by that model are attributable to the company."
In other words: if you consciously decide to rely solely on ChatGPT and forgo a specialist, you cannot later claim that the AI failed to identify the problems.
What this means for the bid process
This ruling is not an argument against using AI in tenders. On the contrary, it shows how important it is to use AI correctly:
Domain specific AI excels at the groundwork. Tender documents often span hundreds of pages. Domain specific AI is trained to identify relevant passages with near 100% accuracy, detect contradictions, and understand the structure of a tender. Generic AI like ChatGPT often achieves only 50 to 70% accuracy.
AI does not replace procurement law assessment. The question of whether a missing maximum quantity in a framework agreement constitutes a procurement law violation requires knowledge of current ECJ case law and German procurement law. Domain specific AI can flag unclear information in tender documents and proactively suggest bidder questions. A final legal assessment should then be handed over to a specialist.
The optimal process combines both. AI handles the review and analysis of documents. A procurement law expert evaluates the results and identifies legal risks. This creates a bid process that is both fast and legally sound.
In this ruling, the Procurement Tribunal explicitly clarified for the first time that companies bear responsibility for the use of AI tools. Those who use AI as a sole substitute for legal advice bear the full risk. Those who use AI as a tool and combine it with expert advice gain a real competitive advantage.
Vergabekammer Niedersachsen, Beschluss vom 28.04.2025, VgK-14/2025
Customer stories
How Arsipa increased its win rate by 78%
Arsipa uses Forgent AI to centralize and scale its tender management across a growing group, achieving 83% less effort in bid decisions and a 78% higher win rate.
Felicitas von Rauch
News
Building AI-Enabled Engineering, Product & Design Teams in 2026
This article shares practical lessons from building Forgent AI's engineering organisation around autonomous coding agents, covering what changed when AI crossed a capability threshold in early 2026.
Leonard Wossnig
News
ChatGPT instead of an expert? The Public Procurement Tribunal says: That's your risk.
A landmark 2025 ruling by the Lower Saxony Procurement Tribunal confirmed that companies bear responsibility when relying on generic AI to review tender documents. The takeaway: combine domain-specific AI with expert knowledge.
Felicitas von Rauch