Legal Risks of the Use of Artificial Intelligence by Firms in Public Procurement and Contracts

Summary

For firms that work on public contracts, artificial intelligence is a useful aid for reading documents, preparing checklists and drafting. In Turkish public procurement and contract law, however, an action is more than producing text: time limits, remedies, the validity of documents, whether a precedent is still current, guarantees and debarment all have to be assessed together.

If firms carry out tender and contract procedures relying only on an AI answer, the result may be exclusion of the bid, a missed deadline for an application, an incorrect transaction in EKAP (the Electronic Public Procurement Platform), forfeiture of a guarantee, termination of the contract, debarment from public tenders, or the breach of trade secrets or personal data. This article examines these legal risks; how AI can be built into a firm's internal workflow is a separate subject.

1. The legal nature of AI output

Under the current definition adopted by the OECD, an AI system is a machine-based system that, for explicit or implicit objectives, infers from the input it receives how to generate outputs such as predictions, content, recommendations or decisions. Generative AI refers to systems that can produce text, tables, summaries and drafts from instructions given in natural language. The Turkish Personal Data Protection Authority's Guide on Generative AI and the Protection of Personal Data also notes that these systems carry ethical, legal and social risks and should be approached in a transparent, auditable and human-centred way.

From the point of view of public procurement and contract law, the key point is this: AI does not take legal responsibility for the action. Its answer does not replace a signed legal opinion, an expert report or a decision by an authorised person in the firm; the firm bears the consequences of the action.

2. Benefits and their limits

Used correctly, AI saves time in summarising documents, putting specification items into tables, separating the provisions on delivery periods, penalties, price adjustment and inspection and acceptance, and preparing a first list of questions. These benefits do not remove the need for legal review at the points that have legal consequences, such as calculating time limits, the right to apply, the validity of documents, whether decisions are still current, termination, guarantees and debarment.

3. Main legal risks

3.1. Applying the current legislation wrongly or incompletely. Public procurement and contract law changes frequently. Public Procurement Law No. 4734 (Kamu İhale Kanunu) and Public Procurement Contracts Law No. 4735 (Kamu İhale Sözleşmeleri Kanunu) are applied together with the implementing regulations for goods, services, works and consultancy services, the Implementing Regulation on Conducting Public Procurement Electronically, the Public Procurement General Communiqué, the price adjustment rules and the rules on applications. AI may base its answer on an outdated text of a regulation, an electronic tendering practice that is no longer in force, a provision of the Communiqué as it stood before amendment, or an approach of the Public Procurement Board that no longer applies. For example, if a firm acts without checking the current rules on which year's data are used for economic and financial standing documents in a tender held in the first four months of the year, whether a stamp duty document is required for work experience certificates, or how the rules on explaining abnormally low tenders apply to goods contracts, it may be excluded from the tender.

3.2. Errors in time limits and remedies. In procurement law, time limits often directly determine whether a right can be exercised. The time limits for a complaint against the tender documents, the addendum process, the contracting authority's response, an application to the Public Procurement Authority, bringing a court action, the invitation to sign the contract, submitting a guarantee and explaining an abnormally low tender all differ. A correct calculation is impossible without considering together when the documents were downloaded, whether there is an addendum, whether a bid was submitted, whether a complaint was made to the contracting authority, the date of its response and whether the contract has been signed; AI, however, may set a wrong time limit based on incomplete dates given by the user. A general answer such as "a complaint can be made", "an application to the Authority is possible" or "a court action can be brought" is not enough: an application made at the wrong stage or by the wrong route may be rejected on grounds of time limit, standing, jurisdiction or form without the merits being examined. The remedy and the time limit should be assessed legally against the actual calendar.

3.3. Non-existent or incorrect precedents. One of the main weaknesses of AI in law is that it may produce decision numbers that do not exist, wrong court decisions or inaccurate summaries of Board decisions. An assessment published by Stanford HAI found that general-purpose chatbots produce "hallucinations" at high rates in response to legal questions and that inaccurate output has not been eliminated even in some legal-specific tools. In public procurement law, a precedent has meaning together with the type of tender, the subject of the dispute, the date of the decision, the legislation applied, the Board's reasoning, whether it has been reviewed by the courts and the specific structure of the documents. A complaint, application or statement of claim based on an incorrect precedent weakens the case. Presenting a decision that does not exist as if it were real undermines the credibility of the application and, depending on the circumstances, may raise further questions of liability. Every decision number given by AI should be verified against the official source.

3.4. Reading the documents only as text. AI can summarise tender documents, but the problems in them often arise not from a single sentence but from the combined effect of several provisions. When the delivery period in the technical specification, the delay penalty in the draft contract, the qualification document in the administrative specification and the inspection and acceptance provisions are read together, the risk the firm will take on becomes visible; in some cases a delivery period provision may in practice prevent participation. Assessing the documents in terms of competition, equal treatment, qualification, bid costs, delivery periods, technical feasibility, penalties and post-contract risks cannot be left to AI.

3.5. EKAP and electronic tendering. The date of entry into force of the Implementing Regulation on Conducting Public Procurement Electronically was set as 1 August 2025 by an announcement of the Public Procurement Authority dated 26 December 2024. With this transition, electronic tendering, e-signatures, EKAP records, pre-contract documents, confirmations and electronic applications have become more important. Without seeing the EKAP screen or knowing the firm's actual registration status, AI may give instructions such as "upload this document", "enter this information in this field" or "submit the guarantee this way". The e-signature holder, the authority to represent and bind the firm, the amount of the guarantee, the Certificate of Participation in the Tender (İhaleye Katılım Belgesi), the time at which documents are uploaded, the tax and social security debt checks, submitting an explanation of an abnormally low tender and submitting pre-contract documents are concrete actions with legal consequences.

3.6. Explaining an abnormally low tender. An explanation of an abnormally low tender is not just a well-written text produced by AI. It is built on the significant components of the bid, cost elements, price quotes, supporting documents, analyses, labour cost calculations, material costs, transport, depreciation, compliance with the technical specification and the dates of the documents. The process is carried out under Law No. 4734 and the Public Procurement General Communiqué (see "Common Mistakes in Querying Abnormally Low Tenders in Service Procurement and How to Address Them", in Turkish). An explanation prepared by AI should not be uploaded directly to EKAP: consistency between the text and the documents, the completeness of the significant cost components and the conformity of the explanation with the practice of the Public Procurement Board and the Council of State (Danıştay) should be checked separately.

3.7. Qualification documents and work experience. Qualification documents consist of many items, including evidence of economic and financial standing and of professional and technical capacity (work experience, quality and standard certificates, authorisation certificates and documents required by specific legislation). Even if AI names a document correctly, it may not always assess correctly whether it can be requested at the bidding stage, the contract stage or the inspection and acceptance stage. Work experience is too broad a subject to be left to a one-sentence answer.

3.8. Non-price criteria and assessment of competition. In some tenders, non-price criteria may mean that even the lowest bidder does not win. AI can explain the formula mathematically, but without concrete knowledge of the market and the legislation it has difficulty assessing whether the formula restricts competition, whether it in practice favours a particular firm, or whether the structure of the documents complies with the basic principles of the Law. In its announcement of 9 April 2026, the Public Procurement Authority explained how the scoring criterion is to be set when the consistency between the bid and the structure of the estimated cost is used as a non-price criterion in works tenders; this subject requires up-to-date knowledge of the legislation and practice.

3.9. Interpreting Law No. 4735 as if it governed a private-law contract. Public Procurement Contracts Law No. 4735 lays down the principles and procedures for drawing up and performing contracts awarded under Law No. 4734. Public procurement contracts are not texts that can be freely amended and reshaped by additional protocols like private-law contracts. AI, however, may try to resolve a dispute with private-law reasoning and general statements such as "the price can be changed by an additional protocol" or "the scope of work can be changed by mutual agreement". Time extensions, price adjustment, increases in the scope of work, termination, guarantees and debarment are subject to special rules; failure to carry out the contract process in line with the legislation may lead to sanctions such as termination, debarment and forfeiture of the guarantee. A contract provision should not be applied on the basis of an AI interpretation alone.

3.10. Calculating progress payments, price adjustment and time extensions. Progress payments, price adjustment, time extensions and delay penalties can only be settled by assessing together the dates, the work schedule, quantities produced, indices, the price adjustment decree, increases in scope, the revised work schedule, inspection and acceptance reports, delays caused by the contracting authority and the contractor's written requests. The fact that the price adjustment rules appear under a separate heading on the Public Procurement Authority's legislation page also shows that this area requires specific calculation and legal knowledge. A wrong period, a wrong index, an incomplete application or incorrect documents may lead to the claim being rejected; AI cannot reliably settle the figures of a specific progress payment file and the actions of the contracting authority without review.

3.11. Forfeiture of guarantees and debarment. Some errors do not stop at losing the tender; they may lead to serious consequences such as forfeiture of the bid or performance guarantee, termination of the contract and debarment from public tenders. According to the Public Procurement Authority's 2025 Public Procurement Monitoring Report, there were 7,815 active debarment records as at 31 December 2025, and 6,276 debarment records were entered in 2025. AI may tell the firm to "object", "write a defence" or "plead force majeure", but it cannot assess on the basis of the actual documents on what legal ground the guarantee was forfeited, whether the formal notice was properly served, whether the contractor was at fault, whether a delay caused by the contracting authority is documented, or on which statutory provision the debarment decision is based.

3.12. Personal data and trade secrets. Uploading the bid schedule, price analyses, supplier quotes, staff lists, payroll data, social security and tax documents, draft contracts and correspondence with the contracting authority to public AI systems together with the tender documents may create legal problems. The Personal Data Protection Authority's guide states that personal data processing carried out through generative AI systems must be assessed under Law No. 6698. Price analyses, supply chains, discount rates, labour costs, technical solutions, work schedules and subcontractor information may be trade secrets; uncontrolled sharing may result both in a personal data breach and in disclosure of the content of the bid.

3.13. Unclear responsibility within the firm. When action is taken on the basis of an AI answer, it is often not recorded which document was uploaded, what data the answer was based on, who approved it, who took the decision and who signed the final document. The NIST AI Risk Management Framework provides a structure for managing the risks of uncontrolled and unrecorded use of AI in organisational processes. "That is what the AI said" does not protect the firm from the legal consequences of exclusion, loss of a guarantee or debarment.

3.14. Evidence and correspondence. In tender and contract disputes, being right is often not enough; it has to be proved within the time limit and with the right documents. Requests for time extensions, delays caused by the contracting authority, force majeure, defective performance claims, unpaid progress payments, price adjustment claims, impossibility of delivery, inspection and acceptance disputes and termination proceedings depend on written evidence. AI can provide an impressive draft petition, but it cannot keep track by itself of which record must be made, which letter must be sent on which date, how an objection must be raised, how a notice by registered e-mail (KEP), through the court system (UYAP) or through EKAP must be documented, or how the site diary or the inspection and acceptance report should be used.

4. Assessment

AI is not a tool that should be excluded entirely from public procurement and contract processes; used correctly, it brings speed, order, drafting support and early awareness. For firms working on public contracts, however, the main risk is not so much that AI gives wrong information as that it presents wrong information in confident, well-organised text. What matters in public procurement law is the correct time limit, the correct document, the correct remedy, the correct precedent and correct contract management. AI can be used in preparation; every step with consequences should be assessed on the basis of the actual documents from a legal, financial and technical point of view.

Related article: for the internal workflow and control points of AI use, see Kılıç Kurumsal, "Artificial Intelligence in Public Procurement and Contract Processes: Internal Workflow and Control Points" (14 October 2026).

İlyas Kılıç, Public Administration Specialist