Story perspectives

Anthropic Adds Claude Watermark, Releases EU Compliance API

8/17/2026

Summary

Anthropic Watermarks Claude

Anthropic announced on August 15, 2026 that Claude will embed a watermark for EU AI Act compliance.
The watermark creates a pattern in token choices that readers cannot see but a detection key can identify.
Claude will add a C2PA provenance tag to files, with Anthropic releasing a detection API and the watermark applied to models released on or after August 2, 2026, while older models will be updated later.

Summary

Anthropic Service Outage

Anthropic shut down Claude services—including claude.ai, Console, API, Code, and Cowork—from 21:58 UTC to 22:34 UTC on August 16, 2026, a 36-minute outage.
It was the fourteenth incident logged in two weeks, after outages on August 12-15 and earlier in the month.
Users reported sign-in failures, blank screens and requests, prompting BleepingComputer to record an error page on claude.ai.
Frequent degradations have led enterprise developers to adopt multi-provider fallback routing for production reliability.

History

Historical Context

After the EU enforced the GDPR on 25 May 2018, fines of up to €20 million or 4 % of global turnover pushed Google, Facebook and other tech giants to embed privacy safeguards. Building on that regulatory pressure, Anthropic introduced a covert watermark and C2PA provenance tag for Claude in August 2026, forcing enterprise users to adopt multi-provider fallback strategies.

1 of 3

Anthropic Watermarks Claude

Anthropic announced on August 15, 2026 that Claude will embed a watermark for EU AI Act compliance.
The watermark creates a pattern in token choices that readers cannot see but a detection key can identify.
Claude will add a C2PA provenance tag to files, with Anthropic releasing a detection API and the watermark applied to models released on or after August 2, 2026, while older models will be updated later.

1 / 3