AI governance

Responsible AI

How Real Talk Studio governs AI in practice sessions — built on ISO 42001 and designed to keep learners safe.

Real Talk Studio uses AI to power realistic roleplay — but realism must not come at the cost of safety. Our ISO 42001:2023 certification reflects a structured approach to identifying AI risks, governing how models are used, and maintaining accountability as the product evolves.

ISO 42001

What it means in practice

  • Risk-based design — AI features are assessed for impact before release, with controls matched to how learners interact with avatars during high-stakes scenarios.
  • Lifecycle governance — Model selection, prompt design, guardrails, and monitoring are documented and reviewed as part of our AI management system.
  • Transparency — Learners practice in clearly defined scenarios. AI avatars stay in role; they do not impersonate real individuals without explicit scenario design.
  • Accountability — Issues are tracked, reviewed, and fed back into product and policy updates — not treated as one-off incidents.

Certificate No. 517972026 · Registered by Citation ISO Certification

User safety

How we keep users safe

Controls designed for practice under pressure — where generic chatbot guardrails are not enough.

Session guardrails

Live moderation and topic guardrails run during practice. Content that crosses safe boundaries triggers warnings or session termination — designed for conflict-training scenarios without allowing harm.

Scenario boundaries

Simulations are scoped to defined learning objectives. Avatars respond within the scenario context rather than open-ended general chat, reducing unpredictable behaviour.

Post-session moderation

Completed sessions are reviewed automatically to support ongoing safety monitoring and provide an audit trail for compliance teams.

Human oversight

Administrators and L&D teams configure scenarios, review completion evidence, and can align practice to internal conduct standards and policies.

Data minimisation

We collect what is needed to deliver and evidence practice sessions. Customer simulation content is not used to train foundation models.

Enterprise controls

Organisation and studio permissions, credit governance, and transcript access are structured for teams that need separation of duties and audit-ready records.

Formal policy document

For legal and compliance review, our Responsible AI Policy sets out contractual commitments and usage guidelines in full.

Read Responsible AI Policy

Questions about AI governance for your rollout? Contact us or visit the Trust Centre.