Employees with concealable stigmatized identities (e.g., LGBTQ+ status, neurodivergence, invisible disabilities) choose how much to disclose at work, and firms decide how strongly to invest in inclusion programs. Both decisions are influenced by a shifting external social climate that neither party controls. I develop a two-period principal--agent model of this relationship. Identity disclosure cannot be fully retracted, as it persists in the organization's memory. I show that the firm may strategically provide high inclusion early on to incentivize disclosure, and then cut inclusion supports later. This allows the firm to extract surplus from the employee who is locked in by what is already known about her. Exploitation of this disclosure persistence can be optimal for the firm even in stable or improving social climates, when one would not expect disclosure irreversibility to matter. I also show that more organizational memory is not always worse for employees: the organization's memory can serve as an implicit commitment device, although the firm's preferred level of memory is always at least as high as the employee's.