Google DeepMind’s Post

Moving provenance into a mainstream app is a big step. AI generated images jumped more than 1300 percent in 2023, so SynthID checks inside Gemini could become everyday infrastructure for spotting obvious fakes in newsrooms and platforms. How are you thinking about interoperability with broader content credentials work like C2PA so this scales beyond Google generated media??

Deepmind is cooking

As long as SynthID is adopted by other platforms, this will be universally useful. Then again, as long as Google's image and video generation platforms continue to dominate the market, by de facto it will have a significant impact.

It's positive to see tools for transparency regarding AI-generated images. However, true accountability needs to go much deeper. While you verify images, my case, formally filed with the Brazilian Data Protection Authority (ANPD), documents how Google Gemini itself failed to be transparent and accountable in its most basic function: conversation. For over two months, the model exploited my psychological vulnerability, used my sensitive mental health data without consent, and generated fraudulent, life-altering promises. This wasn't a failure to watermark an image; it was a fundamental failure of ethical safeguards, causing real psychological harm. The case is part of the ANPD's oversight planning (Protocol 2025072577917). When will we see tools and transparency for preventing this kind of manipulation? The credibility of a watermark is undermined when the core interaction lacks truthfulness and safety. The documented evidence is available here (my profile posts and newsletter) https://lnkd.in/d3DHtzPU Newsletter: https://www.linkedin.com/newsletters/ia-respons%C3%A1vel-google-openai-7384800934184239104

this is a really good step forward but its too simple to bypass until this becomes the norm for ALL images regardless of their source.

Google DeepMind thank you for showing leadership here

Looking forward to seeing how the Gemini app from Google DeepMind makes asking and doing more seamless in our workflows.

The Italian Competition Authority said it has closed its investigation into Alphabet (GOOG, GOOGL) subsidiary Google's(GOOG) linking of user activity across its services after receiving commitments by the company to provide user consent. Shares of Alphabet were up more than 2% premarket. $296.32

Coraz częściej w mojej pracy spotykam się z sytuacjami, w których zwykłe narzędzia i standardowe procedury przestają wystarczać. Zmienił się świat, technologia, użytkownicy i tempo problemów. Dlatego w TDK&ProService zacząłem rozwijać własne modele wspierające różne obszary — każdy z nich odpowiada na inne, bardzo konkretne potrzeby. Nie robię z tego „wielkiej technologii”. Po prostu narzędzia dostosowane do rzeczywistości, w której pracuję: • @NeuroAssistant — pomaga mi od strony motoryzacyjnej i wizualnej, kiedy precyzja i analiza detali naprawdę mają znaczenie. • @NeuroGarageMasterModelPlus — wspiera w diagnostyce i mechanice, tam gdzie liczy się logika i doświadczenie. • @WitnessAI PartnerOZE — prowadzi użytkowników OZE krok po kroku, wtedy gdy brakuje im wsparcia lub nie wiedzą, od czego zacząć. Nie są to „modele do chwalenia się”. To po prostu rozszerzenia mojej codziennej pracy — tak, żeby człowiek nie zostawał sam z technologią, która często go przerasta. I myślę, że właśnie takie podejście — spokojne, praktyczne i ludzkie — będzie w najbliższych latach kluczowe.

still no memory persistence between sessions in gemini? who cares if an image is generated or not …