AI News AI资讯 7d ago Updated 7d ago 更新于 7天前 52

You can now turn off Google Gemini's visible watermarks 你现在可以关闭Google Gemini的可见水印了

Google is making visible watermarks optional in AI-generated images, videos, and music through Gemini and Flow Invisible SynthID watermarks and C2PA metadata will remain embedded regardless of the toggle setting The change aligns Google with competitors like OpenAI and Meta, who already rely on invisible watermarking standards Visible watermark removal will not be available in countries that legally require them Anthropic also announced invisible watermarking for AI-generated text and images thi Google允许用户在Gemini和Flow中关闭AI生成内容的可见"sparkle"水印 关闭后内容仍保留不可见的SynthID水印和C2PA元数据用于后续验证 其他AI公司(OpenAI、Meta、Anthropic)已采用类似的不可见水印策略 Google计划将此设置扩展到Search,但不会在强制要求可见水印的国家推出

68
Hot 热度
65
Quality 质量
62
Impact 影响力

Analysis 深度分析

TL;DR

  • Google is making visible watermarks optional in AI-generated images, videos, and music through Gemini and Flow
  • Invisible SynthID watermarks and C2PA metadata will remain embedded regardless of the toggle setting
  • The change aligns Google with competitors like OpenAI and Meta, who already rely on invisible watermarking standards
  • Visible watermark removal will not be available in countries that legally require them
  • Anthropic also announced invisible watermarking for AI-generated text and images this week

Why It Matters

This shift signals a major industry pivot from visible to invisible watermarking as the standard for AI content attribution, reflecting growing pressure from users and creators who find visible watermarks intrusive. It also raises important questions about content verification and the potential for misuse when visible markers are removed, making it critical for practitioners to understand the underlying verification technologies.

Technical Details

  • Google's "Media watermark" toggle in Gemini and Flow disables the visible "sparkle" watermark on content generated by Nano Banana and Omni models
  • Invisible SynthID watermarks and C2PA (Coalition for Content Provenance and Authenticity) metadata remain embedded in the background regardless of user settings
  • Verification is still possible through Gemini or Google Search, which can detect these markers and identify AI-generated content
  • The policy excludes countries with legal requirements for visible watermarks, indicating regional compliance considerations in implementation

Industry Insight

  • The industry is converging on invisible watermarking as the preferred model, suggesting visible watermarks are seen as a temporary measure rather than a long-term solution for AI content attribution
  • AI professionals should familiarize themselves with SynthID and C2PA standards, as these will likely become the baseline for content provenance across platforms
  • The removal of visible watermarks increases the importance of verification tools and detection systems, creating both a challenge and an opportunity for companies building AI content authentication solutions

TL;DR

  • Google允许用户在Gemini和Flow中关闭AI生成内容的可见"sparkle"水印
  • 关闭后内容仍保留不可见的SynthID水印和C2PA元数据用于后续验证
  • 其他AI公司(OpenAI、Meta、Anthropic)已采用类似的不可见水印策略
  • Google计划将此设置扩展到Search,但不会在强制要求可见水印的国家推出

为什么值得看

Google此举标志着AI内容水印策略从"可见标识"向"隐形验证"转变,与行业主流做法趋同,反映了AI生成内容标识的标准化趋势。对从业者而言,理解这一转变有助于把握AI内容溯源技术的演进方向。

技术解析

  • Google的Nano Banana和Omni模型生成的内容将不再强制显示可见水印,用户可通过Gemini设置中的"Media watermark"选项关闭
  • 不可见的SynthID水印和C2PA元数据仍会嵌入生成内容中,用于后续验证AI生成身份
  • 用户可通过Gemini或Search工具检测内容是否包含这些隐形标记
  • OpenAI同样依赖SynthID和C2PA,Meta推出了Content Seal标准,Anthropic则同时为文本和图像添加隐形水印

行业启示

  • AI内容标识正从"强制可见"转向"隐形嵌入+可验证"模式,行业共识正在形成
  • 这一转变平衡了用户体验与内容溯源需求,但可能增加公众辨别AI内容的难度
  • 各国监管政策差异将影响水印策略的地区性实施,Google已明确表示不会在强制要求可见水印的国家推出此功能

Disclaimer: The above content is generated by AI and is for reference only. 免责声明:以上内容由 AI 生成,仅供参考。

Gemini Gemini Image Generation 图像生成 Video Generation 视频生成 Policy 政策 Ethics 伦理