Anthropic

Uses your conversations to train AI unless you opt out

Your chat history with Claude becomes training data for Anthropic's models by default, even though you likely didn't realize this when signing up.

Claude can take actions on third-party services you authorize

Once you grant Claude access to Gmail, Google Drive, or other services, it can send emails, modify files, or retrieve data on your behalf until you manually disconnect it—creating ongoing privacy risks if you forget to revoke access.

Your data is transferred to the US and trained on without your ongoing consent

Even if you opt out of training, Anthropic still uses your inputs for "safety review" purposes and trains models on them anyway—making the opt-out choice less protective than it appears.