(YouTube’s Advertiser Brand Safety Relies on AI Content Classification.)

The AI looks at many things in a video. It checks the audio, visuals, and text. It also considers comments and titles. Based on this, it assigns a label like “suitable,” “limited,” or “not suitable” for ads. Videos labeled as “not suitable” usually contain harmful or sensitive material. Ads are blocked from these videos automatically.

YouTube says this system works fast and at scale. Millions of videos are uploaded every day. Human review alone cannot handle that volume. So AI steps in to make quick decisions. The company claims its models are always learning and improving. They get better over time by studying new data and feedback.

Advertisers want control over where their messages appear. YouTube gives them tools to pick which content types they allow. These settings work with the AI’s labels. Brands can avoid risky content without slowing down their ad campaigns.

Some creators have raised concerns. They say videos sometimes get mislabeled. A video about serious topics might be marked “not suitable” even if it is educational. YouTube admits mistakes happen. It offers an appeals process. Creators can ask for a human review if they think the AI got it wrong.

(YouTube’s Advertiser Brand Safety Relies on AI Content Classification.)

The system is not perfect. But YouTube believes AI is the best way to balance safety, speed, and scale. It continues to update its technology to reduce errors and support both advertisers and creators.

(evan spiegel)

Previously, the plaintiff had filed a lawsuit against Nvidia, Meta and ByteDance for similar reasons.

The latest proposed class action lawsuit was submitted to the United States District Court for the Central District of California last Friday. The plaintiff specifically pointed out that Snap used a large-scale video language dataset called HD-VILA-100M and other datasets limited to academic research purposes. The plaintiff claims that in order to use the dataset for commercial purposes, Snap circumvented YouTube’s technical restrictions, terms of service, and license provisions prohibiting commercial use.

The lawsuit demands statutory compensation and applies for a permanent injunction to prevent potential infringement in the future.

This case is mainly led by the creators of the h3h3 YouTube channel with a subscription volume of 5.52 million, as well as the smaller golf channels MrShortGame Golf and Golfholics.

This is the latest case among numerous content creators suing AI model suppliers. Previously, there have been copyright disputes from publishers, writers, newspapers, user generated content platforms, artists, and other parties. This is not the first lawsuit initiated by YouTube creators. According to data from the non-profit organization Copyright Alliance, there have been over 70 copyright infringement cases against AI companies.

The progress of such lawsuits varies: in the case of Meta and Writers Group, the judge ruled in favor of tech giants; In the case between Anthropic and the author group, the AI giant chose to settle with the plaintiff and pay compensation. Currently, the majority of cases are still under active trial.

Roger Luo said：This case centers on whether the commercial use of “research-only” datasets for AI training constitutes a substantive violation of both original content copyrights and platform terms of service. It touches on the universal legal challenge in the age of generative AI: defining the boundaries of data ownership and fair use in training materials.

All articles and pictures are from the Internet. If there are any copyright issues, please contact us in time to delete.

Inquiry us [contact-form-7]

(Google YouTube tests “AI Separation”)

YouTube announced a trial today called “AI Separation.” This test aims to clearly label videos altered by artificial intelligence. YouTube says this is crucial to help viewers spot misleading content, especially important news topics. The test starts soon for viewers in the United States watching English videos.

YouTube will place special labels directly on videos. These labels will identify AI-generated content. Viewers might see these labels appear on the video player itself. The goal is increased openness about how videos are made. YouTube believes viewers deserve this information.

The company focuses on realistic AI content. This includes fake events, altered real footage, or synthetic versions of real people. Examples are AI-made scenes showing disasters that never occurred or fake speeches by public figures. YouTube wants viewers to know when they see this kind of content. The labels will explain the AI involvement.

YouTube acknowledges the challenge. AI tools are constantly changing. The company states it will keep improving detection methods. This test is part of YouTube’s larger effort against online deception. Ensuring trust is vital, especially during elections.

(Google YouTube tests “AI Separation”)

This initial test targets a limited group. YouTube will watch how viewers react. Feedback will shape future steps. The test runs for several weeks. YouTube plans to share findings later.