This post is based on Ars Technica’s article“German router maker is latest company to inadvertently clarify the LGPL license”. The article covers the details of the lawsuit between AVM and Sebastian Steck and the importance of LGPL license compliance.

1. Case Overview

On January 9, 2025, the Software Freedom Conservancy (SFC)announced that the lawsuit it had filed against AVM, a German network equipment manufacturer, had concluded. The core of this lawsuit concerned the rights of users specified in the GNU Lesser General Public License (LGPL) version 2.1, particularly the obligation to provide installation information.

Sebastian Steck, a German software developer, purchased an AVM router in May 2021 and discovered that the source code provided by AVM did not allow him to reinstall modified software on the router. Steck demanded that AVM “provide the complete source code and the compilation and installation scripts for the uClibc, libblkid, libexif, and libosip2 libraries.” When AVM failed to remedy this, Steck filed a lawsuit in a Berlin court in July 2023.

As a result of the lawsuit, the German court ordered AVM to pay Steck’s attorney’s fees. AVM decided not to appeal this decision. The ruling specifies the allocation of litigation costs, which reflects the economic value and importance of open source license compliance issues.

Disclaimer:

This post was not written by a legal professional and cannot be used as a legal basis. For specific situations related to license and legal issues, please be sure to seek advice from a legal professional. Also, this post was written based on publicly available information and may not reflect the positions of all parties to the lawsuit. Please refer to the original text for the full content and context of the ruling.

2. Background and Progress of the Lawsuit

Background

In May 2021, German software developer Sebastian Steck purchased AVM’s popularFritz!Box 4020 router. Steck requested the source code used in the firmware of this router, and this is where the problem arose. The source code AVM provided did not allow him to reinstall modified software on the router.

Legal Basis of the Lawsuit

An important characteristic of this lawsuit is that Sebastian Steck was able to file the lawsuit even though he was not the copyright holder of the LGPL-2.1 software. This is because the LGPL-2.1 license has the character of a contract for the benefit of third parties. According to thecomplaint, users also have the right to receive the source code under LGPL-2.1:

“This license agreement represents a genuine contract in favor of third parties in accordance with Section 328 of the German Civil Code (BGB), namely in favor of the users who receive the software in object code and, in accordance with the wording of the LGPL-2.1 license conditions to be handed over to them, have a direct right to the transfer of the complete corresponding source code.”

This legal basis significantly strengthens the rights of open source software users. When a manufacturer fails to properly fulfill its license obligations, not only the copyright holder but also ordinary users can now take legal action.

Progress of the Lawsuit

1. May 7, 2021: Steck requested the source code for firmware version 6.83 of the Fritz!Box 4020 from AVM
2. May 11, 2021: An AVM subsidiary provided a source code download link
3. May 14, 2021: Steck pointed out the incompleteness of the provided source code and demanded a correction
4. January 12, 2023: Steck’s attorney explained the legal situation to AVM and demanded correction of the source code
5. March 9, 2023: Steck additionally requested the source code for firmware version 7.02
6. July 27, 2023: Steck filed a lawsuit in the Berlin Regional Court
7. Several months after the lawsuit was filed: AVM provided Steck with all the source code requested, including “the scripts that control library installation”
8. January 9, 2025: The German court ordered AVM to pay Steck’s attorney’s fees. This decision includes a ruling on litigation costs, and AVM decided not to appeal this decision

The Plaintiff’s Claims

The violations of LGPL-2.1 by AVM that Sebastian Steck raised in the complaint are as follows:

1. Failure to provide compilation and installation scripts: The source code AVM initially provided lacked the compilation and installation scripts required by LGPL-2.1. These scripts are an essential element needed to convert the source code into an executable program and install that program on the device.

2. Failure to provide environment variable information: Environment variable configuration information such as KERNEL_LAYOUT, which is essential to the compilation process, was not provided. Environment variables are important information that configure the environment in which a program runs, and without them the source code cannot be compiled properly.

3. Inability to install permanently: The information AVM provided did not allow permanent installation of the modified library on the Fritz!Box. Steck argued that this violates a core requirement of LGPL-2.1, namely that users must be able to continuously use modified software on their own device.

4. Inability to build a firmware image: The provided source code did not allow building an installable firmware image for the Fritz!Box. A firmware image is a file containing the router’s operating system and functionality, and being unable to build one effectively means the software cannot be modified and reinstalled.

Steck argued that these matters infringe on “user software freedom,” a core principle of the LGPL-2.1 license. According to his argument, merely providing the source code is not enough; users must be able to actually modify the code and reinstall it on the device.

[Note] An excerpt from thecomplaint disclosed by SFC:

“The information required for the reinstallation of the compiled program libraries on the Fritz!Box (“installation script”) was also deliberately withheld from the plaintiff. Upon request, the plaintiff only received information that could be used to load the libraries in dispute into the working memory (RAM). However, this is not a sufficient installation on the Fritz!Box because the copy is only created temporarily, meaning “fleetingly [volatile].” When the Fritz!Box is switched off and restarted, the modified versions of the LGPL 2.1 libraries would no longer be present on the device, and the versions created by the defendant would be used instead. This is diametrically opposed to the purpose of the LGPL-2.1, namely, to be able to customize and reinstall the software.”

3. Summary of the Ruling

The ruling includes a decision on litigation costs and specifies that AVM must pay Steck’s attorney’s fees. This reflects the economic value and importance of open source license compliance issues. AVM decided not to appeal this decision.

4. Implications and Various Interpretations

1. LGPL-2.1’s obligation to provide installation information: The complaint raised the claim that AVM must provide installation information under the LGPL-2.1 license. This includes the scripts or information needed to actually make the source code usable. The ruling does not explicitly include this content, but the decision to burden the defendant with litigation costs during the proceedings reflects the strengthening of user rights and the importance of providing installation information. However, the installation information required in this case appears to be at the level of the installation scripts required under GPL/LGPL v2, not the GPL/LGPL-3.0 level.

2. Strengthening of users’ legal rights: This decision allows open source software users to take legal action directly against license violations. This is expected to strengthen the power of the open source community and further promote license compliance by companies. In particular, the fact that the third-party beneficiary logic was also recognized in Germany is an important development.

3. Improved self-correcting capacity of the open source ecosystem: Since not only copyright holders but also ordinary users can now demand license compliance, the self-correcting capacity of the open source ecosystem is expected to improve significantly. This contributes to maintaining the freedom of open source software and a collaborative development environment.

4. Expanded scope of corporate responsibility: Companies must now fulfill their LGPL-2.1 license obligations not only to copyright holders but to all users. This expands the scope of corporate responsibility and will require more thorough license compliance. This can increase the transparency and reliability of open source software use.

5. Reconfirming the importance of open source license compliance: Companies must more thoroughly comply with the relevant license terms when using open source software. They must recognize that the obligations under LGPL are not light simply because it is LGPL. However, it is difficult to say that this decision brought a major change in the legal interpretation of GPL installation information. AVM voluntarily provided the source code and installation scripts; the court did not issue a ruling on the interpretation of GPL/LGPL.

6. Existence of various interpretations: Various interpretations exist regarding this case. There is also a view that the Software Freedom Conservancy (SFC) somewhat overstates the significance of this case, because the ruling itself does not present a clear judgment on the interpretation of GPL/LGPL. These various interpretations show that the discussion on open source licenses and user rights is still ongoing.

Considering these implications and various interpretations, companies should improve their internal processes for open source license compliance and respond to user demands more quickly and proactively. Similar cases in the future are expected to further clarify these issues.

This post is based on JBB Rechtsanwält:innen’s blog post “To Mine or Not To Mine” (https://jbb.de/to-mine-or-not-to-mine/) and is published to explain a recent German court ruling on text and data mining (TDM) and to share related knowledge.

Please note that I am not a legal professional, and this content cannot serve as a legal basis. For specific situations related to license and legal issues, please be sure to seek advice from a legal professional.

Background

In 2021, German photographer Robert Kneschke learned that his photos had been included without authorization in an AI training dataset created by the nonprofit organizationLAION (Large-scale Artificial Intelligence Open Network).

An AI training dataset refers to a large collection of data used to train artificial intelligence models. The dataset called ‘LAION-5B’ consisted of about 5.8 billion images and their corresponding description text. Such datasets are used to improve an AI’s ability to recognize and understand images.

CommonCrawl

At the heart of this case is the nonprofit organization ‘CommonCrawl’, which plays an important role. CommonCrawl regularly creates a ‘backup’ or ‘snapshot’ of the internet. It replicates, in text form, every webpage accessible through links.

• How CommonCrawl collects data:
  1. It replicates the text content of webpages.
  2. It does not directly store non-text data such as images or videos.
  3. Instead, it stores the source code of webpages, which includes links to such content.

CommonCrawlmakes the datasets it collects available on its own website. This dataset includes the ‘source code’ of webpages, which researchers can use to analyze the structure and content of the internet.

LAION’s Data Processing

LAION used this dataset provided by CommonCrawl tocreate its own image dataset. This process is as follows:

1. Extracting image links from the CommonCrawl dataset: LAION filtered the CommonCrawl data to find only the links to image files.

2. Collecting additional information: LAION sought to collect not only image links but also additional information about each image. This additional informationincludes:

   • Image description
   • Presence of a watermark
   • Whether the image contains content harmful to minors

3. Downloading and analyzing images: To obtain this additional information, LAION downloaded the actual images through the collected links and analyzed the images using its own AI models.

4. Constructing the dataset: The final dataset LAION created was structured as a table, with each row containing an image link and additional information about the corresponding image.

Through this process, LAION built a large-scale image dataset that could be used for AI training. However, copyright issues were raised during this process, which eventually led to a legal dispute.

Kneschke argued that even though the terms of service of the website containing his photo prohibited automated content downloading, LAION’s unauthorized downloading and analysis of his photo constituted copyright infringement. In response, LAION countered that its activities fell under text and data mining (TDM) for scientific research purposes and were permitted under Section 60d of the Copyright Act.

This case raised important legal and ethical questions about how to strike a balance between data collection and copyright protection in the AI era.

The Start of the Lawsuit

On April 27, 2023, Kneschke filed a copyright infringement lawsuit against LAION in the Hamburg Regional Court. Copyright infringement refers to the use of a copyrighted work without the copyright holder’s permission. Kneschke objected to the unauthorized use of his photo and demanded that his image be removed from the dataset. This raised an important question about how to protect creators’ rights in the AI era.

Legal Issues

The core issues of this lawsuit are as follows:

1. The scope of application of the text and data mining (TDM) exception: The TDM exception refers to a provision in copyright law that allows a copyrighted work to be used without the copyright holder’s permission under certain conditions. This applies when large volumes of data need to be analyzed for research or technological development. In this lawsuit, the issue was whether creating a dataset for AI training falls under this exception. For example, it had to be determined whether automatically collecting and analyzing a website’s text for research purposes constitutes copyright infringement, or whether it falls under this exception and is permitted.
2. The definition of noncommercial scientific research purposes: The issue was exactly what LAION’s claimed ’noncommercial scientific research’ means, and whether its activities fall under this definition.
3. The validity of the copyright holder’s ‘opt-out’ right: ‘Opt-out’ refers to the right of a copyright holder to refuse to have their work used for TDM. The issue was how this right can be exercised and what form of refusal is valid.

The Impact of the EU Copyright Directive

In 2019, the EU adopted the Digital Single Market Copyright Directive (DSM Directive), which came into effect in EU member states starting June 7, 2021. This directive included two exceptions for text and data mining:

1. TDM for scientific research purposes (Article 3)
   • Scope: Applies only to research organizations and cultural heritage institutions.
   • Purpose: Permitted only for the purpose of scientific research.
   • Authorization: No prior permission from the copyright holder is required, and no compensation of any kind is required.
   • Access condition: Applies only to data that can be legally accessed (e.g., subscriptions, licenses, free online content, etc.)
   • Restriction: Excludes institutions under the decisive influence of private companies.
2. TDM for general purposes (Article 4)
   • Scope: Applies to all individuals or organizations.
   • Purpose: Applies to TDM for any purpose (including commercial purposes).
   • Authorization: Applies only if the copyright holder has not explicitly reserved their rights.
   • Access condition: Applies only to data that can be legally accessed.
     • Opt-out mechanism: The copyright holder can reserve their rights in an ‘appropriate manner’ (e.g., in a machine-readable format for online content).
   • Data retention: Copies may be retained for TDM purposes.

Germany incorporated this directive into domestic law and amended its Copyright Act as follows:

• Section 44b: Established a new exception for TDM for general purposes. This provision permits TDM for any purpose, including commercial purposes, but recognizes the copyright holder’s right to explicitly opt out.
• Section 60d: Expanded the existing exception for TDM for scientific research purposes. This provision grants broader freedom for TDM for noncommercial scientific research purposes and does not recognize the copyright holder’s opt-out right.

The Ruling

On September 27, 2024, the Hamburg Regional Court ruled that LAION’s conduct did not constitute copyright infringement. The main points of the ruling are as follows:

1. LAION’s dataset creation activity falls under TDM for noncommercial scientific research purposes under Section 60d of the German Copyright Act.
2. The mere fact that LAION has a cooperative relationship with commercial companies does not negate its noncommercial nature.
3. A TDM prohibition phrase written in natural language in a website’s terms of service can also be regarded as an opt-out in a ‘machine-readable format’.

Significance of the Ruling

1. A broad interpretation of the TDM exception:
   • The court recognized LAION’s image dataset construction activity as TDM for noncommercial scientific research purposes.
   • This means that modern research methods, such as building AI training datasets, can also fall under the TDM exception.
   • This interpretation could provide greater freedom for AI research and development.
2. An expanded definition of noncommercial research:
   • The court determined that the fact that LAION has a cooperative relationship with commercial companies does not negate its noncommercial nature.
   • This could strengthen legal protection for collaborative research between academia and industry.
   • Not only pure academic research but also industry-academia collaboration projects can now benefit from the TDM exception.
3. A new interpretation of the opt-out mechanism: Although the opt-out did not apply in this case because LAION’s activity was recognized as TDM for noncommercial scientific research purposes, this determination carries important meaning in a broader context:
   • Flexibility of legal interpretation: The court flexibly interpreted the requirement of a ‘machine-readable format’ in line with technological developments. This shows that the law can adapt to a rapidly changing technological environment.
   • Impact on future commercial TDM: Although not applied in this case, this interpretation could carry significant meaning for commercial TDM, because a copyright holder’s opt-out is valid for commercial TDM.
   • Guidance for copyright holders: This ruling provides guidance to copyright holders that, if they wish to exclude their content from TDM, they can specify this clearly in their website’s terms of service.
   • Impact on technology companies: AI and data mining companies may now need to review website terms of service more carefully.
4. Balance between copyright law and technological innovation:
   • This ruling can be seen as an attempt to strike a balance between copyright protection and promoting technological innovation.
   • It provided the legal space needed for the advancement of AI and data science, without completely disregarding the copyright holder’s rights.

Future Outlook

Kneschke can appeal this ruling, and given the importance of the matter, it could go to a higher court or even the Court of Justice of the European Union (CJEU). This ruling is also expected to affect similar cases in other EU member states.

This case raises important legal and ethical questions about how to strike a balance between copyright protection and technological innovation in the AI era. Further discussion and legal judgments in this area are expected to follow.

Implications for Domestic AI Companies

Although this ruling is a German case, it also offers important implications for domestic AI companies:

1. Commercial TDM: While this ruling focuses on noncommercial research, it suggests that commercial TDM may also be permitted under certain conditions. However, for commercial TDM, the copyright holder’s opt-out right must be respected.
2. Data collection methods: AI companies must carefully check a website’s terms of service when collecting data. If a provision explicitly prohibits TDM, this may need to be respected.
3. Research collaboration: Companies could consider building datasets through collaboration with nonprofit research institutions. This could be a way to secure the necessary data while reducing legal risk.
4. Transparency and ethics: It is important to maintain transparency about data use in the AI model development process and to establish ethical guidelines. This can help prevent potential legal disputes.
5. Preparing for domestic legal amendments: Laws similar to the EU Copyright Directive may also be discussed domestically. AI companies need to review their data collection and use policies in advance and adjust them as necessary to prepare for such legal changes.

This case raises important legal and ethical questions about how to strike a balance between copyright protection and technological innovation in the AI era. Domestic AI companies should also keep an eye on this global trend and continue their efforts toward responsible AI development.

As the use of open source software has spread widely, the legal issues surrounding it have grown increasingly complex. In particular, the question of copyright over derivative works based on open source projects that use a copyleft license such as GPL (GNU General Public License) is a thorny subject for many companies. A recent software copyright infringement lawsuit in China offers important implications for this issue.

Parties to the Lawsuit

• Plaintiff: Wangjing Technology (Wangjing)
• Defendants:
  • Yibang Communication Technology (Yibang)
  • Qi’ao Network Technology (Qi’ao)
  • and three individuals (Liu, Wu, Xie)

Overview of the Case

In 2009, Wangjing developed a converged communication smart gateway product called “OfficeTen.”

OfficeTen SDG 1800 by Wangjing -http://www.cncr-it.com/product_detail.php?sid=26&cid=133&id=388

The “OfficeTen1800” software embedded in this product was developed based on the open source framework “OpenWRT,” and obtained a copyright registration certificate from the National Copyright Administration in 2013.

This software consisted of two components: the base system software built on OpenWRT and the upper-layer application software. Wangjing claimed that the latter was an “independent and separate program” from the OpenWRT system.

In 2015, Wangjing began an investigation after suspecting that a competitor, Yibang’s product infringed its copyright. The investigation found that former Wangjing employees had provided the source code of “OfficeTen1800” to Qi’ao, helping it develop very similar software, and that this software was used in Yibang’s product.

According to the appraisal, the proportion of identical non-open-source code between Wangjing’s “OfficeTen1800” and the software used in Yibang’s product reached 90.2%, and Wangjing’s special marks were found in Yibang’s product.

Progress of the Lawsuit

In July 2018, Wangjing filed a software copyright infringement lawsuit against Yibang and Qi’ao. Wangjing demanded that the infringement be stopped and sought damages of 3 million yuan.

The Defendants’ Arguments

Yibang and Qi’ao denied the infringement and argued as follows:

1. “OfficeTen1800” was developed based on the open source framework “OpenWRT.”
2. “OpenWRT” is subject to the constraints of the GPLv2 license.
3. Wangjing’s failure to disclose the source code of “OfficeTen1800” was a violation of GPLv2.
4. Therefore, Wangjing cannot claim copyright over the software.

The Court’s Ruling

First-Instance Judgment

The Suzhou Intermediate People’s Court ruled as follows:

1. Even where a developer modified or made secondary development of an open source product, if it created an original work, it holds copyright in that work.
2. It cannot be concluded that all related software must be disclosed under the GPLv2 agreement.

Accordingly, the court found Yibang and Qi’ao liable for infringement and ordered them to stop the infringement and pay damages of 500,000 yuan (about $70,961, roughly KRW 1 billion).

The Supreme People’s Court’s Ruling

Yibang and Qi’ao appealed, but the Supreme People’s Court upheld the original judgment. The Supreme People’s Court’s main findings were as follows:

1. Since the parties in this case are not the rights holders of the “OpenWRT” system software, whether GPLv2 was complied with cannot be examined in this proceeding.
2. Whether Wangjing violated the GPLv2 agreement and its claim for damages for copyright infringement are separate matters.
3. The copyright arising from a software developer’s original contribution must not be unreasonably deprived or restricted.

Significance of the Ruling

This ruling offers important implications for the copyright protection of derivative works based on open source software.

1. Recognition of Originality: The court held that even a derivative work based on open source software can be subject to copyright protection if the developer made an original contribution.
2. Separation of License Violation from Copyright Protection: The court treated the question of GPLv2 license violation and the claim for damages for copyright infringement as separate matters. This means that even if there is a license violation, the copyright itself can still be valid.
3. Prevention of Rights Abuse: By rejecting the defendants’ argument that “it’s fine to copy it since there’s an obligation to disclose source anyway,” the court prevented reckless copying that abuses the GPL license.
4. Protection of the Open Source Ecosystem: By recognizing copyright in derivative works, the ruling encourages open-source-based innovation and promotes the healthy development of the open source ecosystem.

Similarity to the WordPress Theme Case

In the Karlsruhe Higher Regional Court’s WordPress theme case (ruling of November 13, 2020, reference number 6 U 60/20), GPLv2 was likewise raised as a defense. In that case, the court made the following important findings:

1. A distinction must be made based on whether the copyright holder of the (alleged) derivative work licensed that work under GPLv2.
2. The mere possibility of a copyleft violation is not sufficient to defeat a copyright claim.
3. Enforcement of GPLv2 is the licensor’s responsibility, and it cannot be enforced merely because a user declares the software to be “GPL licensed.”
4. The copyleft effect does not automatically lead to GPL licensing. This is an act that the author of the derivative work must actively carry out.

This finding aligns with the ruling of China’s Supreme People’s Court, and shows a converging trend in the international legal interpretation of GPL licenses and the rights to derivative works.

Implications for Corporate Open Source Management

This ruling offers the following important implications for corporate open source managers:

1. Thorough License Compliance: When using open source software under a copyleft license such as GPL, the requirements of that license must be thoroughly complied with.
2. Importance of Original Contribution: Even when developing based on an open source project, it is important to clearly identify and document original contributions.
3. Source Code Management: Open source code and in-house developed code must be clearly separated and managed.
4. Legal Risk Assessment: Legal risks that may arise from using open source should be assessed and prepared for in advance.
5. Continuous Monitoring: The similarity between a company’s own products and competitors’ products should be continuously monitored to detect potential copyright infringement early.

Conclusion

This ruling from the Chinese court, together with a similar ruling from a German court, clearly resolves the misconception that “GPL-based software products already have an obligation to disclose source anyway, so isn’t it fine to copy them?” Even a derivative work based on open source software under the GPL license can be subject to copyright protection if the developer made an original contribution.

This can be seen as a balanced approach that encourages innovation using open source software while preventing reckless copying and copyright infringement. Companies should refer to this legal interpretation when establishing their open source policies, and strike a balance between license compliance and original development.

As the use of open source software becomes even more common, this kind of legal judgment is expected to be referenced in more countries going forward. Corporate open source managers should therefore continuously monitor these legal trends and reflect them in their own open source policies.

Finally, this ruling delivers an important message to both the open source community and commercial users. It reminds us once again that respecting the spirit of open source while recognizing developers’ effort and creativity, and pursuing innovation while complying with licenses, is the path to a healthy software ecosystem.

References

1. 2024-09-20 OpenWRT, the GPL and the Supreme People’s Court of China:https://www.ifross.org/?q=node/1676
2. 2023-12-29 Copyright dispute cases over derivative works based on open source code:https://www.copyright.or.kr/information-materials/trend/International-copyright-center/download.do?brdctsno=52544&brdctsfileno=22493