Anthropic shares more details about how Claude’s new watermarks will work | TechCrunch · AuraTracer

SummaryAI generated

On August 15, 2026, Anthropic revealed the specific mechanism of the new watermark technology for the Claude model. Previously, on August 12, users had expressed dissatisfaction, believing that this watermark function could potentially lead to accidental detection of usage traces in work and classroom environments, thereby causing related controversies. Currently, Anthropic is further explaining the details of this technology in response to public concerns.

Related eventsRELATED EVENTS

Quick factsQUICK FACTS

SynthID Text proposed by Google DeepMind in 2024Technical source

Transparency code of the EU’s Artificial Intelligence ActName of the regulation

Key entitiesKEY ENTITIES

AnthropicClaudeEU AI ActGoogle DeepMindSynthID

Event frameEVENT FRAME

Regulation

回溯归档across 2 days

Status

Anthropic releases details on the new watermark technology for Claude in response to EU legislation requirements and user concerns

Coverage · reports per dayLANGUAGE SPLIT

Entity relations

Integrated timelineUNIFIED TIMELINE

2026-08-12
Users are dissatisfied with the new watermark technology causing controversy

Anthropic decides to insert invisible code watermarks in outputs to meet EU AI Act requirements; some users believe this mechanism may lead to accidental detection of usage traces in work and classroom environments.
2026-08-15
Anthropic explains the new watermark mechanism for Claude

A blog post explains the use of Google DeepMind SynthID Text technology, with plans to launch a detection API to comply with EU AI Act requirements.

SignalsSIGNALS

Keyword heat

Anthropic2
EU AI Act2
Claude1
SynthID1
Google DeepMind1

All reports (2)SOURCES

↗

T TechCrunch 归档 8月 p11 en 2026-08-12 08:00

Some Claude users are mad that Anthropic's new watermarks will catch them using it at their jobs, classes | TechCrunch

Anthropic 近日决定在 Claude 输出中插入不可见代码水印以标记 AI 生成内容，此举旨在满足欧盟《人工智能法案》透明度准则要求。该政策引发部分用户不满，认为其可能通过数字水印导致学生在作业、记者总结文章等场景中被发现使用 AI。Reddit 上有用户批评该机制对普通用户构成“数字纹身”，并质疑将 Chatbot 视为工具却要求其承担生成责任的不合理性；另有观点指出此举存在双重标准，且未充分考虑用户实际工作流中非直接复制粘贴的情况。