SynthID

2026-10-31

MODELS · 1 SOURCE · The Decoder

Anthropic rolls out watermark to four Claude models, but two older models can be bypassed by changing sampling settings

Anthropic is adding its imperceptible text watermark to Claude Mythos 5, Claude Opus 4.7, Claude Opus 4.6, and Claude Sonnet 4.6; the watermark is embedded at the model level and will appear across hosting platforms (including Amazon Bedrock). On Opus 4.6 and Sonnet 4.6 the watermark only activates when developers leave sampling parameters (temperature, top_p, top_k) at their defaults, so tweaking any one parameter can evade the mark; Anthropic offers a private-preview detection API and cites the EU AI Act deadline as the reason for the rollout.

8.0

2026-10-15

COMPANIES · 1 SOURCE · Google

TIME names Google's SynthID to Best Inventions of 2026

TIME has included Google SynthID on its 2026 Best Inventions list. Google says SynthID invisibly watermarks AI-generated images, audio, video, and text at creation, has tagged over 180 billion images and videos and 240,000 years of audio, and is integrated with Google Search, the Gemini app, Chrome, and a SynthID Detector; Google also announced SynthID Bio for synthetic biology.

6.0

2026-10-07

COMPANIES · 3 SOURCES · Google · TechCrunch AI · The Decoder

Google makes SynthID Detector available globally in English

Google (Google DeepMind) has expanded access to its SynthID Detector, making the tool available globally in English to let anyone check whether images, video, or audio were generated with AI by Google or partner providers (OpenAI, NVIDIA, Kakao, and soon Apple). The company says it has watermarked over 180 billion images and videos and 240,000 years of audio, and that SynthID is integrated into Search, the Gemini app, and Chrome, which together now handle over 1 million verification requests daily.

7.0

2026-10-05

REGULATION · 3 SOURCES · The Decoder · The Verge AI · TechCrunch AI

OpenAI will add invisible text watermarking (textGrain) for ChatGPT in the EU; API opt-in worldwide

OpenAI is rolling out textGrain, an invisible statistical watermarking system, for ChatGPT and Codex users in the European Union to comply with the EU AI Act, and will let API customers worldwide opt in to watermarking (including via cloud partners such as Microsoft Azure). OpenAI says textGrain performed competitively in internal tests, will be open-sourced, but detection reliability varies by text length and topic and degrades substantially with edits, and initial access to the detector will be limited to selected researchers and organizations.

8.0

2026-09-24

MODELS · 1 SOURCE · The Verge AI

Google’s Gemini 3.8 Live gains real-time 'Live Avatar' (enterprise only)

Google added a 'Live Avatar' feature to Gemini 3.8 Live that renders an animated persona which lip-syncs and shows facial expressions in near real-time; the capability is currently available only to Gemini Enterprise customers. Google says the avatar can transition across the 97 supported languages without visual drift, surface on‑screen information while speaking, and outputs include an invisible SynthID watermark; organizations can choose preset avatars or create custom ones.

6.0

2026-09-24

MODELS · 3 SOURCES · Google · Google DeepMind · The Decoder

Google releases Gemini 3.8 Flash and Flash‑Lite text‑to‑speech models

Google introduced two new Gemini 3.8 text‑to‑speech models—Flash TTS for highly expressive, character-driven voice design and Flash‑Lite TTS for cost‑efficient, high‑volume use—available across Google AI Studio, Gemini API/Enterprise, Gemini Notebook, and Google Vids. The models claim large voice coverage (2,000+ production voices, 100+ languages/dialects), generative voice creation from prompts, 30‑second voice replication with consent verification, SynthID watermarking and C2PA credentials, and features for long‑form and multi‑speaker scene staging.

8.0

2026-09-23

COMPANIES · 1 SOURCE · Google

Google adds Gemini Omni 1.1 Flash to Google Vids for free HD AI video creation

Google integrated its Gemini Omni 1.1 Flash model into Google Vids, allowing anyone with a Google or Google Workspace account to generate and edit 1080p HD AI video scenes at no cost. New creative controls include scene extension, exact-duration setting, upscaling, and an imperceptible SynthID watermark; Gemini 3.8 Flash-Lite TTS support is coming soon.

8.0

2026-09-17

RESEARCH · 1 SOURCE · Ars Technica

Research finds SynthID-Text watermarking can alter LLM refusal behavior and tool use

New research by Andrea Siposova of Lasso Security shows that SynthID-Text watermarking (the Google-origin method Anthropic plans to use for Claude) can change not only token selection but also whether models refuse harmful prompts and which tools agents invoke, especially under prompt-injection attacks. The experiments used Hugging Face’s SynthIDTextWatermarkLogitsProcessor on several open-weight models; the study did not test Anthropic’s Claude implementation and notes behavior varied by secret key and model.

7.0