IP Tech Knowledgy

Website owners have tools to help deter artificial intelligence from scraping information and content from their website. Deploying a variety of tactics could create stronger legal arguments against the AI companies that collect your creative work and data to train their systems, and may even reduce the chance that invasive systems would use your content. Contract, copyright, and other laws may provide methods to stop otherwise-undeterred AI systems, or at least provide remedies due to AI taking information from a website without permission. However, successfully bringing a legal action could be time-consuming, difficult, and expensive.

Contractually, various Terms of Use, End User License Agreements, Clickwrap Agreements, or even notices in footers, should include language specifically stating that the license given to use the website and the materials on a website does not include, and specifically prohibits, the right to scrape or copy any information or data. The more specific about what types of usage, file access, copying, etc., is prohibited, the stronger those terms will be, such as mentioning in the license/notice prohibitions on model training, text-and-data mining, dataset creation, and derivative AI uses from the license. Similarly, a narrow list of what types of usage is allowed by the user helps frame the permission and helps in strengthening future legal claims. It is also better to have the website user take some physical action to assent to the licensing terms before being granted access.

Copyright law protects against third parties copying or making derivative works of creative works of authorship, which could be words, images, selected data or computer code. Although no notice on the website is required to exercise copyright protection, adding a clear and conspicuous notice to a website is helpful. Registering for a copyright for the content on a website also allows for enhanced damages, and the ability to collect attorneys’ fees and statutorily prescribed damages unavailable without registration. Obtaining a copyright registration has become more popular since one successful class action case – where the infringing AI trained from online books -- allowed compensation in the settlement only to book publishers who had registered the copyrights in their books.

AI scraping could rise to a criminal act under the Computer Fraud and Abuse Act when the activity evades certain access protections or restrictions or falsifies the actor gaining access.

Implementing technical measures to restrict improper or excessive access also helps. Consider using robots, instructions, rate limits, anti-bot tools, particular IP blocking, geofencing, login tests, and session duration control to stop unwanted access, and adding fake content or watermarks to trap infringers.

Heeding one or more of these suggestions can provide decent legal arguments against a would-be scraper, and should act as a deterrent. Demand letters to the invader, noting the violations of the provided license, could stop the AI intrusion. However, an aggressive AI entity may not heed the warning or the law and may try to take as much as it can.  An individual company might be hesitant to take on the fight and thus unable to protect itself. Class action lawsuits might be a better route to address overbearing attacks on websites.

Ned T. Himmelrich
410-576-4171 • nhimmelrich@gfrlaw.com