False Positives?
What should you do if you receive a letter or email from the BKA stating the following: “that your servers contain content relevant to criminal law,” and "requesting that you remove it as soon as possible"?
Netzpolitik.org reports on an incident involving the internet service provider Flokinet. The operators checked the specified URLs and found that one link leads to a German composer, who produces music for films, and the second video contains a song by a Spanish opera singer. Furthermore, Flokinet merely hosts the Invidious software and mirrors YouTube videos without consuming their data — and thus cannot be held responsible for their content.
The operators responded to the BKA immediately within two hours, pointing out that, in their view, the content was not punishable by law. They received no reply. The BKA’s demand for “immediate deletion” typically entails a 24-hour deadline in many countries; in the case of terrorist content, providers must delete it within one hour.
In response to an inquiry from Netzpolitik.org, the BKA acknowledged that, in isolated cases, links to non-criminal content may have been automatically forwarded. This is likely due to the volume of reports. Many reports come from the hotlines in the INHOPE network, are reviewed by the police, and then forwarded to the BKA. In 2024, there were 31,000 reports. The U.S. organization “National Center for Missing and Exploited Children” reported over 205,000 tips in 2024. However, the BKA found “criminal relevance” in only half of the cases.
Flokinet criticizes that even though the number of reports is so high, every individual case must be given attention, and responses from the suspects should at least be addressed as well. After all, the authorities can only minimize their sources of error if they get to the bottom of false reports. This is particularly true if there is reason to fear that AI programs are involved in the detection process — either now or in the future. If the number of these false positives exceeds single-digit percentages, it can be assumed that the police are automatically reporting content without verification, thereby effectively outsourcing the verification work to the providers. In the reports from the U.S. this seems not to be the case because the BKA deletes about 50% of the messages without relevance.
Translated with DeepL.com (free version)
Category[21]: Unsere Themen in der Presse Short-Link to this page: a-fsa.de/e/3Qd
Link to this page: https://www.a-fsa.de/de/articles/9562-20260613-fehler-automatisierter-erkennung.htm
Link with Tor: http://a6pdp5vmmw4zm5tifrc3qo2pyz7mvnk4zzimpesnckvzinubzmioddad.onion/de/articles/9562-20260613-fehler-automatisierter-erkennung.htm
Tags: #AI #KI #BKA #Polizei #Internet #Überwachung #Kontrolle #Porno #Mißbrauch #Kinder #Sperrung #unverzüglich #FalsePositives #Youtube #Inhope #Invidious
Created: 2026-06-13 09:17:54
Kommentar abgeben




