Researcher Calls Anthropic AI Watermarks Dead On Arrival For EU Rules
Reactions from ranked influencers
2 posts
within hours? How about year ago or more? Had @WIRED journalists done basic literature review before writing, they would’ve written a different story — a story about how out of (technology) touch EU AI regulators are, and how Anthropic researchers knew this “safety” feature was…
Anthropic announced last week it would include invisible watermarks in AI-generated content to comply with new EU rules. Within hours, overrides were being touted online.
Deepmind researchers, who originally published the idea behind this under the name SynthID, knew the limitations of their approach and made the sensible decision to not tout it as some valuable safety feature. Of course, the broad strokes idea predates Deepmind in the LLM…