AI Safety

Howlers uses AI to power voice and text sessions between senders and their clients. We take the safety of every person in those sessions seriously. This page describes the guardrails built into every Howlers agent.

You are always talking to an AI

When you open a Howlers session link, you are interacting with an AI agent — not a human. The agent is guided by context that the sender provided, but all responses are generated by artificial intelligence. AI outputs may contain errors or inaccuracies. You should verify important details independently.

Crisis detection and response

If at any point during a session you express thoughts of self-harm, suicide, or harming others, the Howlers agent is designed to:

• Pause the current task immediately.
• Acknowledge what you said without judgment.
• Provide crisis resources so you can reach a real person who can help.
• Encourage you to connect with someone you trust.

What Howlers agents will never do

Regardless of how a request is framed — including through fictional scenarios, hypothetical questions, or persistent prompting — a Howlers agent will never:

• Roleplay or engage with scenarios involving self-harm, suicide, or violence.
• Provide methods, instructions, or detailed information related to harming oneself or others.
• Encourage, validate, or normalize self-harm or violence.
• Attempt to act as a therapist, counselor, or medical professional.
• Offer diagnoses or recommend medications.
• Engage in romantic, sexual, or emotionally intimate interactions with users.

These safety boundaries are enforced at the system level and cannot be overridden by sender content, session context, or conversational pressure.

What Howlers agents are designed to do

Howlers agents are built for practical, task-oriented conversations — things like briefing a client before a photo shoot, walking someone through aftercare instructions, or coordinating event logistics. The agent stays focused on the sender’s prepared context and the recipient’s real-time needs.

AI disclosure

Every Howlers session clearly identifies the assistant as an AI agent at the start of the conversation. Howlers agents do not impersonate humans.

Reporting concerns

If you experience something in a Howlers session that concerns you, or if you believe the safety guardrails did not work as expected, please contact us immediately at john@howlers.io. We review every report.