From Tuesday, old models of Claude also hide a signature in what they write Anthropogenic sent an email to customers: on November 3 he begins to mark text from Claude myth 5, opposite 4.7, opposite 4.6 and I am yes 4.6. Newer models like Memol 5.5 already carry the same mark. You can't see it. This is a word selection pattern, with no hidden characters. It travels with copy-paste and might survive some editing. Anthropic says it doesn't contain anything about you, your company or your chats. This applies worldwide, not just in the European Union, whose AI law has made it necessary. For now only regulators, police, media, fact checkers and investigators can check it. A hit is not proof that Claude wrote the text. Entropy's help page says that only text that Claude has proofread or translated can carry the mark.

From Tuesday, old models of Claude also hide a signature in what they write Anthropogenic sent an email to customers: on November 3 he begins to mark text from Claude myth 5, opposite 4.7, opposite 4.6 and I am yes 4.6. Newer models like Memol 5.5 already carry the same mark. You can't see it. This is a word selection pattern, with no hidden characters. It travels with copy-paste and might survive some editing. Anthropic says it doesn't contain anything about you, your company or your chats. This applies worldwide, not just in the European Union, whose AI law has made it necessary. For now only regulators, police, media, fact checkers and investigators can check it. A hit is not proof that Claude wrote the text. Entropy's help page says that only text that Claude has proofread or translated can carry the mark.