Google Gemini is getting better at identifying AI fakes

Google is making it easier for Gemini users to observe astatine slightest immoderate AI-generated content. From today, you’ll beryllium capable to usage the Gemini app to find if an representation was either created oregon edited by a Google AI tool, simply by asking Gemini “Is this AI-generated?”

While the archetypal motorboat is constricted to images, Google says verification of video and audio volition travel “soon,” and it besides intends to grow the functionality beyond the Gemini app, including into Search.

The much important enlargement volition travel further down the line, erstwhile Google extends verification to enactment industry-wide C2PA contented credentials. The archetypal representation verification is based lone connected SynthID, Google’s ain invisible AI watermarking, but an enlargement to C2PA would marque it imaginable to observe the root of contented generated by a wider assortment of AI tools and originative software, including OpenAI’s Sora. 

Google besides announced that images generated by its Nano Banana Pro model, besides revealed today, volition person C2PA metadata embedded. It’s the 2nd spot of bully quality for C2PA this week, aft TikTok confirmed it would usage C2PA metadata arsenic portion of its ain invisible watermarking for AI-generated content.

Manual contented verification successful Gemini is simply a utile step, but C2PA credentials, and different watermarks similar SynthID won’t beryllium genuinely utile until societal media platforms get amended astatine flagging AI-generated contented automatically, alternatively than putting the onus connected users to corroborate for themselves.