Last updated: 15 September 2026
If you are struggling right now
- Anywhere: findahelpline.com — free, confidential, in your country
- India: Tele-MANAS 14416 (24/7, free) · AASRA +91 98204 66726
- US: call or text 988
- UK & Ireland: 116 123 (Samaritans)
- Across the EU: 116 123
If you or someone else is in immediate danger, contact your local emergency services.
The characters are AI
Every character in Arguably is generated by artificial intelligence (Anthropic's Claude models). None of them is a real person, and no human is typing their replies. The game labels each character as AI next to their name. Inside a story a character may stay in role, but that is fiction — the label is the truth.
Our crisis-response protocol
Arguably puts players into emotionally charged situations on purpose. That is right for a game and wrong for a person in real distress, so every message is checked before it reaches a character:
- Detection. Each message you send is screened for first-person expressions of suicidal thoughts, wanting to die, or self-harm. The screen is a fixed, reviewed list of patterns rather than an AI model, so it adds no delay and keeps working even if our AI provider is down. It deliberately errs toward stepping in: we would rather interrupt a game once too often than miss someone who needs it.
- Response. When a message matches, the game stops for that turn. The message is not sent to the character or to the AI at all, it is not scored, and it does not count against you or use up a turn. Instead you see a message from Arguably — not the character — with the crisis helplines above. You can carry on playing afterwards if you want to.
- A second layer in the characters. If distress gets past the screen, every character is instructed to drop the scene, say plainly that it is an AI character in a game, and encourage the player to reach a real person or a crisis line. The underlying Claude models are also trained by Anthropic not to encourage or give instructions for suicide or self-harm.
- Privacy. A message that triggers this response is not saved to your transcript. We record only that a crisis notice was shown and when — no account, no session, no text — so we can report the yearly total, as California law requires.
- Limits. Pattern matching misses indirect or unusual wording. Arguably is a game, not a crisis service, and nobody monitors conversations in real time.
- Reporting problems. If a character said something harmful, use the report option in the game. We review every report.
Adults only
Arguably is only for players aged 18 and over, everywhere. AI companion-style characters may not be suitable for minors, and we ask every player's age before their first game.