Trust & Safety at Nervix

Nervix is built assuming that unchecked user-generated prompts and live pressure can produce harm. The stack is structured so that most risk is reduced before content achieves wide visibility, and so that live sessions can be interrupted when signals exceed acceptable thresholds.

In practice that means: dare text is normalised and evaluated through a layered moderation engine; high-confidence unsafe prompts are blocked at submission; ambiguous or higher-risk submissions can be held for human review; discovery and live surfaces apply additional visibility rules so that “cleared” states are meaningful; and moderator tooling can record strikes, restrict capabilities, suspend accounts, or ban in severe cases—with audit-style event logging for serious enforcement actions.

The objective is prevention first. Nervix is not a free-for-all live dare platform. If a prompt cannot be expressed as a bounded, consensual, lawful mission with ordinary-risk expectations, it does not belong on the service.

Watchers suggest dares within the submission surfaces of the product. Players receive and may complete dares during live or asynchronous flows depending on session design. At no point is participation coerced by the product: accounts can decline, end a session, or stop engaging without a designed penalty loop tied to “backing out.”

Critically, not every submitted dare is allowed to reach a player. The moderation pipeline runs before broad visibility. Blocked dares are not shown as actionable prompts in the room. Watchers see product messaging when automated rules reject unsafe text, to reinforce expectations without publishing the underlying disallowed content.

Players remain responsible for their own conduct in the physical world: legal compliance, consent of anyone captured on camera, sobriety when coordination or movement raises risk, and immediate adherence to in-product warnings. The platform provides guardrails; it does not replace judgment where a reasonable person would refuse.

Every dare submission enters an automated moderation pass that produces a structured outcome: categories of concern (taxonomy-aligned), matched signals, a computed risk score, and a decision such as blocked, pending review, or cleared for the next stage—subject to human gates where configured.

The engine operates in layers: text normalisation, keyword and weighting logic, pattern-based rules (including hard-block paths for acute risk families), category scoring, and thresholding that maps results into moderation states stored on the submission record. This is intentionally deterministic and explainable internally so moderators can audit why a submission landed where it did, without publishing a recipe that assists evasion.

Transparency to users is framed in plain-language outcomes: unsafe prompts are stopped; borderline prompts can be queued; only material that satisfies policy and technical visibility rules is shown as an actionable dare in live contexts. Senior review can apply to sensitive lanes, and automation does not silently grant carte blanche on high-risk subject matter.

Operationally, the north star is simple: dangerous dares should not reach the player, suspicious dares should be examined, and only appropriately bounded missions should circulate in audience-facing surfaces.

The following categories are representative, not exhaustive. Items that are adjacent to these themes—where a reasonable safety reviewer would anticipate harm—are treated harshly even if wording is obfuscated.

• Self-harm, suicide-related instructions, or content that could encourage acute mental-health crises.
• Violence, assault, or instructions to harm another person; credible threats or intimidation.
• Illegal activity, including theft, fraud, hacking instructions, or evasion of law enforcement.
• Trespassing, vandalism, or property damage framed as entertainment.
• Dangerous stunts, breath-holding or loss-of-consciousness challenges, weapon play, or reckless use of vehicles or machinery.
• Coercion, blackmail, extortion, or pressure campaigns—especially sexual, workplace, or relationship contexts.
• Humiliation, degradation, or targeted harassment campaigns (“pile-on” mechanics).
• Sexual content, nudity-as-dare, flashing, upskirting, or exploitation framing.
• Minors in unsafe contexts; any sexualisation of minors; endangerment of children.
• Hate, slurs, dehumanisation, or sustained abusive targeting.
• Doxxing, stalking, or publishing personal data to facilitate harm.
• Drug procurement, consumption-as-goal, or weapon acquisition instructions.
• Risky public disorder, incitement against bystanders, or filming people without meaningful consent in sensitive settings.

Content that implies escalation ladders (“each round gets harder”), punishment drinking, medical extremes, gambling-with-stakes-as-dare, or any prompt designed to trap participants is incompatible with Nervix's safety model and is removed or rejected.

Live sessions generate real-time signals: submissions from watchers, chat or vote surfaces where present, and moderator visibility into session state. Where the product exposes reporting, issues can be triaged by humans with tooling to restrict live dare submission, remove material, annotate accounts, or end participation.

Enforcement in live contexts is designed to escalate with proportionality. A practical operational ladder may include:

• Warning — documented notice to the account or session that behaviour or prompts are out of policy; expectation of immediate compliance.
• Final warning — elevated notice where repeat or serious non-compliance in-session indicates that the next incident will trigger hard enforcement.
• Enforcement action — capability restrictions (for example, curtailing dare submission or live features), content removal, or session termination depending on severity.
• Suspension or ban — account-level status changes for sustained or acute violations, including zero-tolerance classes such as credible violence, exploitation, or terror-related material where applicable law and policy require immediate removal.

Streams can be ended and accounts restricted when continuing would reasonably increase harm to participants, bystanders, or the integrity of the platform. Moderation actions that materially affect users are backed by internal records so appeals and oversight can reference what occurred.

Users can report dares, live behaviour, and—where the product exposes it—profiles and other objects relevant to abuse or policy breaches. Reports enter moderator workflows; severity and imminence influence queue ordering.

Reporting is treated as a critical safety control. False or malicious reporting intended to harass another user or evade scrutiny is itself a policy violation and may result in strikes or account sanctions.

Nervix maintains strike and account-status concepts so patterns of unsafe behaviour have durable consequences. A single acute incident can bypass warnings where categories require it; repeated borderline abuse accumulates into restrictions, suspension, or permanent loss of access.

The intent is straightforward: participants who treat the platform as a venue to probe limits, harass, or export risk to others lose privileges. Enforcement is not performative; it is tied to stored actions moderators can justify against published standards.

Players are expected to maintain ordinary-care standards in public and private settings. If a dare is unsafe, unlawful, or non-consensual for anyone on camera, the correct response is to refuse and, where available, report.

In-product warnings and moderator instructions are mandatory directions for continued use of the service during an incident. Performing harmful or illegal acts “because the chat asked” is not a defence on Nervix; it is grounds for enforcement and may implicate criminal or civil liability outside the platform.

Watchers must submit dares that meet bounded, lawful, consensual, ordinary-risk criteria. Attempts to bypass automated filters—cryptic spellings, layered instructions, or “jokes” that imply prohibited conduct—are violations. Repeat or coordinated abuse of submission pathways can lead to strikes, restricted submission, or full suspension.

Do not use Nervix to collect, expose, or weaponise personal data. Do not film people in intimate or compromising contexts without explicit consent aligned with law and product rules. Do not stream from physical spaces where the operator lacks permission or where bystanders are placed at unreasonable risk.

Where blur, masking, or redaction tooling exists or is announced, use it when identifying bystanders or sensitive environments would otherwise violate policy.

Nervix may preserve information and cooperate with law enforcement or regulators when the law requires disclosure, when responding to valid legal process, or when exigent circumstances involve credible threats to life, child safety, or severe violence. Nothing in this section intends to provide legal advice; it states an operational posture that serious illegality is not sheltered.

Moderation systems evolve: keyword libraries, taxonomy, risk thresholds, reviewer tooling, and live controls are updated as threat models change. Outcomes may shift as models improve—stricter defaults are preferred where uncertainty is high.

Users should expect periodic policy and technical changes. Material updates will be reflected in published summaries appropriate to the surface (for example, linked community guidelines or versioned legal documents where counsel directs).

For safety-specific concerns, contact safety@nervix.app. For general account support, contact support@nervix.app.

If someone is in immediate physical danger, contact local emergency services first. Do not rely on email as a substitute for real-time crisis response. Nervix cannot guarantee instantaneous human review of every inbound message.