Key Takeaways
- Google launches a portal to detect AI-generated media
- It scans uploaded files for AI-generated content
- It only detects content watermarked with Google’s SynthID technology
Google Launches New AI Detection Portal
Google has announced the SynthID Detector Portal, a new online service that reveals when files have been generated or edited with AI.
The site allows you to upload files and scan them for the imperceptible SynthID watermarks Google automatically embeds in media generated with the company’s AI tools. It also highlights which areas within an image are likely modified by AI, allowing you to identify specific AI-powered changes within otherwise genuine photos.
Google recently added SynthID detection to Google Photos, allowing users to see when the company’s Magic Editor has been used to manipulate a photo.
The new Detector Portal will enable anyone to perform similar checks on a broader range of media, including text and images generated with Gemini, video generated with Veo and audio generated with Lyria.
Google’s SynthID Detector Portal — How It Works
Google’s AI tools automatically embed invisible SynthID watermarks when creating AI-generated media.
These watermarks are designed to withstand basic digital manipulation. Simple edits won’t erase them, and they remain detectable even when shared online or via messaging apps.
The SynthID Detector portal alerts users whenever it finds a SynthID watermark embedded in an uploaded file.
Google’s SynthID Detector Portal — Serious Limitations
The SynthID Detector Portal is an important first step towards protecting users from AI-based misinformation, such as deepfakes, and helping users differentiate original work from AI-generated synthetic media.
However, It cannot detect content from platforms like ChatGPT or other tools that don’t use SynthID.
Google is, however, developing partnerships with third parties, including Nvidia in March of this year, to help expand the use of SynthID outside of Google. Today, the company announced a new partnership with content verification company GetReal, which will add SynthID detection to its existing array of verification tools.
Conclusion
Watermarks can only form part of the solution in the fight against deepfakes and AI-based deception. Open-source AI tools will always exist, and these can’t be compelled to embed identifying watermarks, making their output more challenging to detect.
For now, you may feel AI-generated content is easy to spot, but that won’t always be the case. As synthetic media becomes more sophisticated, tools like SynthID will become crucial in differentiating it from human-created work.
Follow @paul_monckton on Instagram.
