Google DeepMind announced that SynthID Detector is now available worldwide. The web service lets users upload an image, audio or video file and get a simple yes‑or‑no answer about whether the file contains a watermark from any of the SynthID partners. At launch the partners are Google, OpenAI, NVIDIA and Kakao, with Apple support promised later.
How the detector works and its limits
SynthID embeds an invisible cryptographic watermark into AI‑generated content at the time of creation. The detector reads that watermark using a secret key. Google notes that a “no confident signal” result does not guarantee the content is human‑made – the watermark can be lost after heavy editing or if the creator uses a different watermarking system, such as those from Meta or Microsoft. False negatives are possible, and the tool only covers the listed partners.
Why it matters
For everyday users, the detector offers a quick way to verify the provenance of images, audio or video before sharing them online. It reduces the need to juggle multiple brand‑specific detectors, though it does not replace a broader verification strategy for content from other AI sources.
Why it matters
People can now check a single file for a watermark from several major AI generators, which helps avoid accidental sharing of synthetic media. However, the detector won’t work for content from other providers, so users still need extra tools for a full verification.
