Arwen Protect uses multiple AI techniques to achieves 96–99% detection accuracy across key threat types, including targeted harassment, threats and coordinated abuse.

Alerts are confidence-scored and context-aware, so teams see credible risk, rather than noise. This helps teams act early without overreacting.

Yes. Control is built in.

What one organisation treats as critical may be informational for another, and Protect reflects that. During onboarding we define your Protect policy, setting out what threats you want detected, who needs to be notified, how investigations should be handled and protocols for escalation and security liaison.

We routinely review this policy with you and continually optimise how Arwen is configured to deliver it.

When Protect identifies a credible risk, it follows a clear, predefined workflow:

• The threat is flagged and supporting context and evidence captured

• It is moderated, if that's part of your policy

• Investigation is undertaken, either by us or your in-house team

• Risk assessment and recommended next steps are surfaced

• Escalation paths are triggered if and when thresholds are met

Teams remain in control, with clear documentation at every stage.

Yes. Arwen Protect is built for security- and compliance-conscious environments.

Access is permissioned, investigations are logged and all actions are auditable. Data is handled in line with applicable privacy and security standards and investigation outputs are designed to be shared safely with internal or external stakeholders when needed.

All processing happens via official APIs with OAUTH2 access, no login credentials stored.
Protect is ISO 27001 certified, GDPR compliant, and supports regulator-ready reporting for online safety, legal and governance teams.

Moderation focuses on removing unwanted content. Arwen Protect focuses on identifying and managing real risk.

Protect looks beyond individual comments to understand patterns, intent, escalation signals, and actors. It supports investigation, context-building, and escalation - not just content removal.

However, you can choose to have threatening comments moderated via the Arwen Moderate product.

In short: moderation keeps spaces clean; Protect keeps people and assets safe.

Listening tools tell you what’s being said, usually after the fact.

Arwen Protect operates in real time and is designed for action, not reporting. It connects detection, investigation, escalation, and documentation into a single workflow, so teams can intervene early and responsibly.

OSINT (Open-Source Intelligence) uses publicly available information to help investigate commenters, understand behaviour and identify patterns.

OSINT is used responsibly to add context to high-risk cases - not to surveil broadly. It helps teams distinguish between noise, coordinated activity, and credible threats, supporting informed decisions and safe escalation.

No. It supports them.

Arwen Protect acts as an early-warning and context layer, surfacing risks sooner and packaging evidence clearly. Final decisions and actions remain with your internal teams or trusted partners.

No - and that’s intentional.

Routine detection and prioritisation happen automatically, at speed. Judgement, escalation and intervention remain human-led.

Arwen Protect is designed to reduce cognitive load, not remove accountability.

All actions in Arwen Protect are transparent, reviewable and reversible.

Teams can adjust thresholds, provide feedback, and refine Arwen's activity over time. In practice, customers see fewer errors than with manual review and significantly less risk from missed escalation.

Arwen Protect is most often owned by:

• Trust & Safety teams

• Risk and Reputation teams

• Security and safeguarding teams

Senior leadership typically has visibility through clear reporting and audit trails, without needing to manage day-to-day operations.

Most teams see value immediately through early risk detection and reduced manual effort.

The longer-term value comes from avoided incidents, calmer escalation, defensible decision-making and the deterrent effect of prosecutions - outcomes that are difficult to achieve with reactive or fragmented tools.

Yes.

Protect is designed for early warning, not constant escalation. Many customers use it precisely so that low-level signals don’t quietly grow into major incidents.

Many clients think of it as an insurance policy.

It’s about preparedness, not panic.