Resolves YES if, by 11:59pm ET on June 30, 2027, it is credibly publicly confirmed that Anthropic has made consent to training on user inputs and/or outputs a *requirement* for accessing any frontier-tier model, such that a user cannot use that model without permitting their data to be used for training. A default-on setting that users can opt out of does NOT count. The consent must be required with no opt-out for at least one frontier-tier model. "Frontier-tier" means a flagship Anthropic model at or near the capability frontier at the time (e.g. a Claude Opus or Sonnet class model), not a legacy or deprecated model. The requirement need only apply to a single frontier-tier model or access path, not to all of them. Acceptable confirmation: - Anthropic's official terms of service, usage policy, product documentation, or an official announcement establishing the requirement; or - Reporting from a major outlet (e.g. Bloomberg, Reuters, WSJ, NYT, FT, The Verge, Washington Post) establishing it. Resolves NO if, by that date, no such requirement is confirmed, including if Anthropic continues to offer frontier-tier access with an opt-out, never trains on user data, or the status remains unconfirmed. Absence of evidence resolves NO. **Notes for adjudication:** - "Training" includes pre-training, fine-tuning, RLHF/RLAIF, distillation, or constitutional-AI-style use that updates model weights. A consent requirement scoped only to transient operational use (safety classifiers, abuse monitoring, serving-time steering) does NOT count. - Free or research-preview tiers count as long as the model offered is frontier-tier. A paid tier that still permits opt-out does not trigger YES on its own. - If genuinely ambiguous at the deadline, resolves NO per the default.