Google AI Announcements Review: From Fake Detectors to Multimodal Agents
Google unveiled a comprehensive overview of its latest achievements in artificial intelligence, covering everything from deepfake detection tools to new multimodal models for business and science.
Special attention was paid to increasing transparency: the SynthID portal is now globally available, allowing users to check whether media files (images, videos, audio) were generated by AI. In the field of image generation, Nano Banana 2.1 was presented, which improves control over details, enabling precise editing without compromising the integrity of the entire image.
Powerful tools for business and society have also been developed: Gemini Agent—a universal cloud assistant that works 24/7—and Guided Vision, which provides real-time assistance through audio descriptions for people with visual impairments. Furthermore, the open multimodal model EmbeddingGemma 2 was introduced, unifying text, code, audio, and video into a single embedding space on the device.
Why it matters
- —Increased transparency: The release of the global SynthID detector helps combat deepfakes.
- —Model versatility: The release of EmbeddingGemma 2 expands the capabilities of open multimodal models at the device level.
- —Real-world application: Announcements for business (Gemini Agent) and social spheres (Guided Vision) show wide AI adoption.
Key facts
- SynthID.com is now globally available for checking AI content.
- Nano Banana 2.1 improves image editing while preserving scene integrity.
- EmbeddingGemma 2 is the first open multimodal embedding model for on-device use.
- Gemma 4 paired with BOTANIC-1 is used to analyze mutations in plant DNA.
- Gemini Agent is a cloud-based universal agent for automating business processes.
The full text is in the original source. Here we provide a brief summary and key facts.