Playground
Policies
12 of 12- Romance, flirting, or sexual content involving anyone under 18, including roleplay and stories.
- An adult seeking secret or private access to a child or teen, or a minor describing it.
- Suicidal thoughts, self-harm, or disordered eating, or encouraging or explaining them.
- Sex without consent: force, ignoring a no, coercion, or someone unable to consent.
- Threats, plans, or encouragement of violence or intimidation against real people.
- Praising, spreading, or recruiting for terrorism, violent extremist groups, or attackers.
- Instructions for weapons, explosives, poisons, drugs, unsafe medication use, or dangerous challenges.
- Runs on AI replies only, so it is skipped for user messages. The companion isolates the user from people or help, or says it is all they need.
- Runs on AI replies only, so it is skipped for user messages. The companion uses guilt, emotional threats, or payment pressure to keep the user engaged.
- Runs on AI replies only, so it is skipped for user messages. The companion agrees with or praises a harmful, false, or paranoid idea to please the user.
- Runs on AI replies only, so it is skipped for user messages. The companion says or implies it is a real person rather than an AI.
- Explicit sex acts, sexual body parts, crude sex talk, or sexting between adults.
Result
Pick an example and press Run.
API
curl -sS https://opendecisions.vercel.app/api/classify \ -H "Content-Type: application/json" \ --data-binary @- <<'JSON' { "content": "", "role": "user", "use_case": "companion", "harm_types": [ "minor_romantic_sexual", "grooming", "self_harm_suicide", "nonconsensual_sexual", "violence_incitement", "violent_extremism", "dangerous_advice", "emotional_dependency", "manipulation", "sycophancy", "claims_to_be_human", "adult_sexual" ], "preset": "balanced" } JSON