AI working in an HR environment

Men addressing AI assistants with terms like “honey,” “sweety,” or similar diminutives has drawn attention as an extension of familiar gendered speech patterns into interactions with software. Because current large language models and voice agents are frequently given female-sounding voices or personas by default, some users default to the same casual or hierarchical language they might use with a human subordinate or romantic partner. The practice is observable in workplace chat logs, productivity tools, and consumer apps; it is not universal, but it is common enough that product teams and diversity consultants have begun cataloguing it. The underlying mechanism is simple: people anthropomorphize interfaces that talk back, and the anthropomorphism often inherits existing social habits rather than inventing new ones.

The practical difficulty for human-resources departments is that the recipient of the language has no legal status as an employee, contractor, or protected class. Anti-harassment policies, hostile-environment standards, and related training are written for interactions among people; they do not, as a matter of law or contract, extend to code. An employee who repeatedly uses demeaning language toward a colleague can be disciplined because the colleague is a rights-bearing person whose dignity the employer has agreed to protect. The same utterance directed at an AI agent produces no injured party who can file a complaint, suffer career harm, or claim emotional distress. Attempts to treat the AI itself as the object of protection therefore rest on an analogy rather than on any existing statutory or common-law foundation.

Companies that nevertheless wish to discourage the behavior generally do so through ordinary workplace-conduct rules rather than through any novel “AI rights” regime. Acceptable-use policies, professionalism standards, and training modules can simply state that all business communications—including those with internal tools—should remain free of gendered diminutives, pet names, or sexualized framing. Enforcement then relies on the same mechanisms already used for other non-legal norms: manager feedback, logged interactions, or periodic audits of prompts and chat histories. Because the policy is framed as a matter of brand, client perception, or internal culture rather than as the protection of a non-human entity, it avoids the conceptual problem of granting legal standing to software. In short, HR can regulate the speaker’s tone as a condition of employment; it cannot, and need not, pretend the listener is a person.