Generative AI not only supports the writing of texts and the creation of images but increasingly also the programming of software. This can have an impact on the applicability of copyright protection to software. This is also important in the case of company acquisitions. Where the software has been largely developed by AI, it may lack copyright protection which affects the intrinsic value of the target company. This must be taken into account in the context of due diligence as well as in the drafting of contracts.
Computer programmes are protected by copyright if they represent individual works in the sense that they are the result of their author's own intellectual creation. According to this principle, computer programmes have usually been protected by copyright until now. Since generative AI has also found its way into software development, the question arises as to how this affects the protectability. In essence, a computer programme will be able to protect if a human uses AI only as a subordinate tool. This will be the case if, for instance, AI tests the software to be developed and uncovers inconsistencies. If, however, relevant parts of the code are generated by AI, they will in many cases lack protectability. The exact distinctions are currently still subject to further development. Nevertheless, one should not be too hasty to speak of a gray area overall. The principles are already relatively clear but the specific outcome depends highly on the facts of each individual case.
All this applies regardless of whether the AI provider grants the user (i.e. software developer) all rights to the work results of the AI. If no copyright is created because the human contribution is too small, no copyright can be transferred.
Nowadays, software developers can hardly do without the use of AI in development. The question is thus less whether AI will be used but rather how it will be used. And this question should also be asked as part of the due diligence. Disclosed internal guidelines and documentation on employee training on the use of AI can provide information, as can relevant license agreements. In addition, it is advisable to consult dedicated experts so that the actual use can be verified as accurately as possible. With only superficial due diligence, risks could be overlooked; if W&I insurance is to be taken out as part of the transaction, the associated policy could cancel or reduce the scope of the guarantee to the extent that gaps have been identified in the due diligence.
In addition to general guarantees of ownership of all relevant intellectual property rights (IP), there are separate guarantees with regard to the use of AI, for which attachments with specific descriptions or disclosures may then be manufactured. If the guarantee clause were too generic, there would be a risk that, in the event of an (alleged) breach of the guarantee, legal ambiguity would arise as to whether the particular case falls under the guarantee or not.
A question that must be separated from the above considerations but is nevertheless related, is whether any AI-generated code infringes the rights of third parties. This could be the case, for example, if the AI largely reproduces the foreign code – with which it was trained. The risk can be reduced with a software scan but not completely eliminated; on the other hand, there is also the risk of foreign code being incorporated when human programmers are used. On the other hand, there is likely to be a greater risk if visual content is AI-generated. However, this is not the subject of this article.
If the software solution in question itself represents AI as part of the target company or as an asset to be transferred, not only the AI Act should be kept in mind but also the origin of the datasets with which it was trained. If these are due to web scraping (automated reading of data on websites using software bots or scripts), copyright and data protection questions may arise. Even the acquisition of a license from a commercial provider is only useful if the latter in turn has all the necessary rights himself; a look at the license terms is advisable in any case (for instance, with regard to limits of use by the licensee).
In cross-border company acquisitions, particularly when the target company operates internationally or even globally, the copyright challenges relating to AI-generated software are even greater. Copyright is basically national law, although the requirements for the creation of copyright protection are similar in most countries and thus also the principles for the legal issues relevant here. However, developments are still ongoing at the international level, and a Chinese court may decide the legal issues raised here differently than an American court. The guarantee declarations of the transaction documents should also contain flexible wording that takes into consideration the different legal frameworks in the jurisdictions concerned.
If AI is not only used as a subordinate tool in the creation of software, the software may not be eligible for protection. The software development process should therefore be scrutinized as part of the due diligence process. The findings will have to be reflected in the contract documentation when the company is acquired. Other follow-up issues such as third-party rights, AI Act, scraping and data licenses as well as international aspects must also be considered.
![[Translate to English:]](/fileadmin/_processed_/a/0/csm_IP_Header_Scott_738ee40874.jpeg "[Translate to English:]")
