Transparency
We tell users when AI is involved, what model class is in use at a high level, and where AI output appears. We document AI behavior changes that materially affect users.
Human Oversight
AI output is presented as a recommendation that a human reviews and accepts. We do not design features that send, publish, or commit on a user's behalf without explicit human confirmation.
User Responsibility
Users remain the author of record for any text they publish, send, or submit. We do not represent AI suggestions as a substitute for the user's own judgment.
Continuous Improvement
We evaluate AI features against quality, safety, and bias criteria, and we revise prompts, models, and guardrails as we learn. Material changes are documented.
Responsible Innovation
We decline to build features whose primary use is deception, impersonation, surveillance of individuals, or evasion of legal disclosure obligations, even where such features would be commercially viable.