Is it allowed to train AI with other people’s data?

Web crawlers download foreign data from the Internet every second. This data is then sorted, labeled – and used for AI training. The amazing thing: German law offers far-reaching possibilities in § 44b Act on Copyright and Related Rights (Urheberrechtsgesetz) to legally use foreign data for AI training.

Something different applies if the rights holders have exceptionally reserved the “use for text and data mining”. For works accessible online, such a reservation of use must be in machine-readable form. In practice, most websites and media have not declared an effective reservation of use. We show you the gold standard to maximum protection of your intellectual property.

Who owns the training data? The trained system? And the prompts?

In principle, intellectual property can exist in training data, trained AI system (weights and thresholds) and input (prompts). For this, the specific case must fulfill the legal requirements for protection. In addition, protection under the German Trade Secrets Act (Geschäftsgeheimnisgesetz) can be considered, as long as appropriate secrecy measures have been taken.

Many AI providers obtain a right to reuse all data provided. The AI use is then “paid for” with training or productive data. Especially when cloud tools are used to facilitate work, there is a risk of an (unnoticed) outflow of sensitive company data to third-party servers and to third-party companies. Contractual protection in the service relationship and AI policies for the company’s own employees are indispensable for the protection of trade secrets.

Currently, the best way to protect secrets is to train and operate AI only on your own instances. On premise as the ideal solution. But beware: the AI Liability Directive proposed by the European Commission on 28.09.2022 provides for far-reaching information claims by any AI victims. Therefore, there is a threat of trade secret disclosure.

Aitava shows you solutions on how to comprehensively protect intellectual property and trade secrets.

Who owns AI results?

European copyright law only protects works created by humans. If AI autonomously generates content, this is in the public domain in the copyright sense. They can therefore be freely used and copied by anyone, unless special circumstances apply – as in the German Law on the Protection of Trade Secrets (Geschäftsgeheimnisgesetz).

Liability trap 1: Anyone who promises an exclusive right of use in contracts and delivers public domain AI results is performing deficiently. Developers are well advised to safeguard the creation of intellectual property through well-measured AI use. And document it through AI guidelines.

Liability trap 2: If the AI reproduces (unnoticed) already existing works, the AI output may infringe third party intellectual property rights. Especially with homogeneous training data, this can happen. To reduce this risk, protective measures should be considered, such as automated matching in terms of output control.

Aitava helps you to efficiently secure AI results and avoid liability traps. We see ourselves as an enabler.

We have answers and even more questions.

Exchange at eye level?

Please feel free to contact us for an uncomplicated (virtual) meeting.

We also advise on:

AI & Data Strategy

If you want to be relevant tomorrow, you need to make strategic decisions today. Aitava removes obstacles so that companies can seize the opportunities of AI.

Learn more

AI & Data Compliance

Even today, numerous regulations must be observed when using AI and Data. The AI Act and Data Act herald a new era in technology regulation.

Learn more

Data Sharing

Many companies are sitting on a treasure trove of data. Other companies can make use of this treasure trove of data. We help with data exchange.

Learn more

Liability in the Use of AI

In a world where decisions are increasingly made by AI, the damages caused by AI use are also increasing. We help you identify and mitigate liability risks.

Learn more