YouTube focuses on transparency: this is how content generated by AI will be indicated
It has now become almost impossible not to come across content generated with artificial intelligence and more and more users want to have it greater clarity about. Precisely for this reason, YouTube has made it known that creators will now be required to communicate presence Of ”altered or synthetic elements, including generative artificial intelligence” within your videos. The press release refers exclusively to ”realistic content” that a viewer could easily exchange for a real person, place, scene, or event.
To put this innovation into practice, a new tools in Creator Studio which will allow those who publish the content to include a label in the video description or directly in the player (in this case for more delicate topics such as health, news, elections or finance).
Below are some examples of affected content:
- Using the likeness of a realistic person – digital alteration to replace the face of an individual with that of another or synthetic generation of a person’s voice.
- Altering footage of real events or locations – make it appear as if a real building has caught fire or alter a real cityscape to make it appear different from reality.
- Generation of realistic scenes: viewing a realistic depiction of major fictional events, such as a tornado moving toward a real city.
Instead, it won’t be necessary reveal the use of generative artificial intelligence for generating scripts, content ideas or automatic captions.
They stay excluded Also:
- Clearly unrealistic content
- Color adjustment or lighting filters
- Special effects such as background blur or vintage effects
- Beauty filters or other visual enhancements
Further information and examples on the dedicated support page.
Users will begin to see labels in the next weeks, on all platforms and in all formats. The company concludes by saying that it will give its community the time to adapt to the new features but reserves the right, in the future, to take action towards channels that systematically choose not to report the requested information.