Texts that match one of your account's keywords are no longer checked for an implicit opt-out with AI, so someone texting a keyword that is similar to an opt-out word now won't be unsubscribed by mistake. Explicit opt-outs such as STOP still opt the person out.