Generative artificial intelligence is experiencing a moment of deep contradiction. On one hand, platforms like YouTube are raising the bar for transparency with automatic detection systems for AI-generated content. On the other, tech giants like Google still show embarrassing cracks in their own technologies. Meanwhile, startups like ElevenLabs and Triomics prove that AI can be both a source of creativity and a powerful tool for medicine.
YouTube raises the barrier against invisible AI
The most disruptive news comes from YouTube. The platform announced the introduction of an automatic labeling system for videos generated by artificial intelligence, even when creators fail to declare it manually. The system triggers when content shows significant photorealistic AI use, such as scenes entirely generated by models like Veo or Dream Screen, or when C2PA metadata indicates fully artificial creation. The label appears below the player for long videos and as an overlay for Shorts. The goal is clear: prevent the audience from being deceived by hyperrealistic content. Creators can challenge the label if they believe it is an error, but for content created with YouTube's official tools, the disclosure is permanent. For non-photorealistic content, such as animations or slight alterations, the declaration remains in the expanded description. This move marks a step forward in the fight against misinformation, as discussed in our deep dive into Android 17 and Gemini Intelligence redefining the mobile ecosystem.
Google's embarrassment: its AI cannot spell Google
While YouTube rushes to fix the issue, Google finds itself dealing with another gaffe. According to a TechCrunch report, the company's language model seems unable to correctly write the name "Google." This problem appears almost symbolic: if Google's AI cannot even spell its own company name, how can we trust it for more complex tasks? This episode, documented on TechCrunch, highlights the persistent limitations of language models, which often suffer from hallucinations and trivial errors. A reminder that despite progress, AI is still far from infallible.
ElevenLabs and music that changes genre mid-track
On a positive note, ElevenLabs has unveiled a new music generation model capable of switching genres mid-track. This feature allows creators to regenerate a section of a song without affecting the rest of the piece. It is a leap forward for AI-assisted music production, offering unprecedented granular control. AI-generated music becomes a flexible tool, no longer just a loop generator.
Triomics brings AI to oncology
On the medical front, Triomics has closed a $22 million Series B round led by Battery Ventures to bring oncology-specific AI to cancer centers. The goal is to help doctors interpret complex clinical data and personalize therapies. A virtuous example of how artificial intelligence can truly save lives, far from the scandals of misspelled names.
In this context, the AI landscape shows two faces: on one side the urgency of regulation and transparency, on the other the development of extraordinary capabilities. YouTube has chosen to act, Google has yet to prove it can fix its mistakes, and startups like ElevenLabs and Triomics are paving the way for a future where AI is increasingly invisible but also more accountable.
Sponsored Protocol
