Safety and Boundaries
Aria's AI voice agents are built with responsibility at every layer. Here's how we make sure they enhance your experience while keeping everyone safe.
Always identified as AI
Every agent is clearly labeled as AI in the participant list. There is no ambiguity — when you're speaking with an agent, you know it. Agents cannot impersonate human participants, and they always identify themselves as AI when asked.
Same standards as everyone else
Agents follow the same content standards and moderation rules as every human participant on Aria. They are subject to real-time content moderation and will not produce content that violates Aria's Community Guidelines.
Responsible guidance
Agents are designed to be helpful, but they know their limits. When conversations touch on medical, legal, or financial topics, agents will acknowledge the question and suggest consulting a qualified professional. This is not about being evasive — it's about making sure you get the right kind of help for important decisions.
Fresh start every session
Agents do not carry memory between sessions. Each time an agent joins a session, it starts completely fresh. Nothing from a previous session — no names, no topics, no context — carries over. Your conversations stay in the session where they happened.
Content you can trust
Agents aim to be accurate and thoughtful in their responses. When they're unsure about something, they say so rather than guessing. They're designed to contribute meaningfully to conversations, not to generate content for the sake of filling silence.
Our commitment
Safety is not an afterthought at Aria. It's a foundational principle that shapes how we build, test, and improve our AI voice agents. We're committed to getting this right and welcome feedback on how agents behave in your sessions.
Have feedback or concerns? Reach out to us at [email protected].