Google says SynthID Detector is now available globally in English for image, video and audio checks. The useful boundary: recognising supported AI provenance is not the same as verifying the story a file tells.
What changed
Google’s 29 October update says SynthID Detector is expanding from its earlier media-professional access to global availability in English. The company describes checks for images, videos and audio made with Google tools or participating partners, including NVIDIA, Kakao and Apple. This is a provenance tool rather than a new content-generation model. We checked the announcement, not every account, file format or partner integration.
The question it can help answer
For an editor or marketing team, the first useful question is whether a file carries a supported signal about how it was produced. That is narrower than asking whether a claim is true. A photograph can be genuine and still be attached to the wrong date or place; an AI-created illustration can be openly labelled and used accurately. Keep those editorial questions separate from the detection result.
Do not turn absence into certainty
A negative or inconclusive result should not be treated as proof that a file contains no AI-generated material. The announcement does not establish universal detection of every generator or transformation. Preserve the original asset and the result you received, then ask where the file came from and what it purports to show. Do not attach an invented confidence score or a HubAI accuracy badge to the output.
A practical newsroom or agency check
Try a small set of non-sensitive files whose origins you already know. Include an openly labelled generated asset and an ordinary original. Record the question asked, the result and whether it helped your reviewer. Do not upload a client’s unpublished material until the organisation has reviewed the service’s current upload and data terms. We have not independently benchmarked detection quality or cleared the service for confidential files.
Keep ownership of the final decision
Use provenance alongside source verification, permission and rights checks. A tool result is evidence to examine, not an instruction to publish or accuse someone of deception. If the origin remains unclear, describe that uncertainty or hold the asset back. The cover is an existing conceptual security illustration, not a screenshot of the detector or a sample of its output.
HUBAI VIEWUse a detector result alongside the original file, its source and an editorial check. A negative result is not proof that a file was made without AI.
Buyer decision signal: Google update · 29 October 2026
What to verify next
1Keep the original file and its source.
2Distinguish AI provenance from factual truth.
3Review upload terms before using confidential assets.
4Record uncertainty rather than inventing a detection score.
Read the evidence
Capabilities, availability and prices can change. HubAI keeps analysis separate from the underlying official material.
01Google DeepMind announcement · updated 29 October 2026Open source ↗02SynthID Detector · official serviceOpen source ↗
Products
Products
Products