Follow the latest coverage, related explainers and connected technology stories.
Google has made SynthID.com available for checking images, video, and audio for invisible watermarks used by Google and its partners, including OpenAI and Nvidia, with a daily limit of approximately 10 checks per user. However, the tool does not detect all AI-generated content, as it depends on model providers that use SynthID and does not prevent watermark removal.
Google announced two new text-to-speech models that enable the creation of custom voices, line-by-line performance direction, and multi-speaker dialogues in more than 100 languages. Gemini 3.8 Flash TTS is available to developers through the Gemini API and Google AI Studio, while Flash-Lite targets large-scale audio applications.
Anthropic plans to add an invisible watermark to text generated by Claude models, enabling an estimate of the likelihood that Claude participated in writing it without materially affecting quality or cost. The company says the move is intended to comply with the European Union’s AI Act, with the watermark to launch globally when available.
Within days of Anthropic announcing the activation of invisible watermarks in text produced by Claude, a group of open-source projects and commercial services emerged claiming to bypass these watermarks. However, these claims cannot currently be verified, as Anthropic has not published the watermarking mechanism or a tool to detect it.
Cloudflare launched a developer preview of WebMCP that allows websites to register tools for AI agents to use inside the browser by enabling a single setting, without modifying the code or origin server. The preview includes the Content Credentials and Site MCP Server packages, with the tools executed inside the visitor’s browser.