Anne Buan: Your Tender Is Legally Binding, Regardless of Whether AI Wrote It
Lawyer Anne Buan from CMS Kluge shared the legal pitfalls that arise when AI meets public procurement. From binding obligations to new types of disputes, here is what you need to know.
What happens legally when AI-generated tenders meet a regulatory framework that requires equal treatment, transparency, and verifiability? That was the central question when Anne Buan, partner and lawyer at CMS Kluge, took the stage at Cobrief's breakfast seminar on March 19. With extensive experience in public procurement within IT and technology, on both the supplier and contracting authority side, she provided a sober but important overview of the legal pitfalls that are already beginning to materialize.
More Tenders, but Also More Noise
Anne began by acknowledging the positive aspects: AI lowers the barrier to submitting tenders, which leads to more and often better tenders in competition. There have been far too many competitions that receive only one tender, and here AI is a positive force.
But the medal has a flip side. With more tenders of increasingly similar quality, it becomes significantly harder for buyers to distinguish between suppliers. And for lawyers, it means more tenders with errors, ambiguities, and contradictions, which in turn creates complicated legal assessments around rejection and clarification.
Anne was particularly attentive to the risk from international suppliers who rely blindly on AI-generated tenders without sufficient understanding of Norwegian procurement traditions. It can quickly become, as she put it, "a lot of soup."
Your Tender Is Binding, Period
Perhaps the most important reminder from the presentation was this: regardless of whether AI helped you write the tender, you are legally bound to fulfill the contract. If you face liability because the tender contained errors, you cannot expect to claim damages from the AI provider. Anne urged everyone to check the terms and conditions, because no AI tool takes responsibility for the content it generates.
Internal quality assurance with human involvement is therefore more important than ever.
Evaluation Will Shift Toward the Verifiable
Anne predicted a clear shift in how public buyers will evaluate tenders going forward. As text-based tenders become more similar, the focus will likely shift toward verifiable factors: certifications, accreditations, and documented results. She also envisions greater use of demonstrations, pilots, case assignments, and interviews, meaning that buyers will increasingly want to see suppliers in action rather than relying on text alone.
In addition, she believes we will see stricter qualification requirements and more selection to limit the number of tenders to be evaluated, particularly in negotiated procedures.
New Types of Disputes on the Horizon
As a lawyer, Anne also provided insight into what future disputes may involve. She pointed to several new issues: errors and contradictions in AI-generated tenders leading to rejection, demands that contracting authorities must verify tender content, suppliers breaching contracts because they offered something they cannot deliver, and eventually claims of errors, bias, or hallucination in AI-based evaluations.
A particularly interesting scenario is what happens when a contracting authority sets up automated evaluation but chooses to override the result, and what legal consequences that may have.
Trade Secrets Going Astray
Anne concluded with a warning that Christian Martinsen had also touched on: the risk of trade secrets ending up in open AI models. Suppliers with important trade secrets should think carefully about whether they dare to include this in tenders that could potentially be processed in systems without adequate confidentiality.
Anne Buan is a partner and lawyer at CMS Kluge, specializing in public procurement and IT. She presented at Cobrief's breakfast seminar on March 19, 2026.