Anthropic adds ID checks; chats ordered handed over

Anthropic has started asking some Claude users to verify their identity with a government ID and a live selfie before they can use certain features. (aol.com) The checks are not universal. Anthropic said they are triggered when its systems detect “potentially fraudulent or abusive behavior,” and outside reports said the rollout affects access to some Claude capabilities rather than the whole service. (aol.com, helpnetsecurity.com) Reports on the rollout said Anthropic is using Persona, an identity verification vendor, and asking for a government-issued photo ID plus a live selfie. Anthropic said the verification data is not used for marketing and is limited to verification and compliance purposes. (helpnetsecurity.com, piunikaweb.com) At the same time, a federal judge in Manhattan ruled that documents a fraud defendant generated with Claude were not protected by attorney-client privilege. The February order required Bradley Heppner to turn over 31 Claude-generated documents in his criminal case. (qz.com, courtlistener.com) Judge Jed Rakoff of the Southern District of New York said no attorney-client relationship exists between a user and a public artificial intelligence platform such as Claude. In the February 17 memorandum, he also said Heppner had waived privilege by sharing information with Claude and Anthropic. (yahoo.com, debevoisedatablog.com) The case has spread quickly through the legal industry. Reuters reported that more than a dozen large U.S. law firms responded by warning clients that chatbot conversations could be discoverable in court. (qz.com, claimsjournal.com) Some firms have gone further than warnings. Sher Tremonte told clients that sharing lawyer communications with an artificial intelligence platform may waive privilege, while other firms have pushed clients toward enterprise tools with tighter confidentiality controls. (claimsjournal.com, law.com) The Heppner ruling turned on specific facts. Court filings said Heppner created the Claude materials on his own, not at his lawyers’ direction, after he had received a grand jury subpoena and while he was a target of the investigation. (thesedonaconference.org, chapman.com) Anthropic has not publicly framed the ID checks as a legal-privacy response. Its public explanation, as reported Thursday, is narrower: the company says the step is aimed at abuse and fraud on Claude. (aol.com) Together, the two developments are changing what “private” means on consumer artificial intelligence platforms. One asks some users to prove who they are before they get in; the other shows that what they produce inside the chatbot may later have to come out. (aol.com, qz.com)