Risk Taxonomy

Risk Families

• Misinformation or hallucinated summary.
• Missing material context.
• Misleading framing or false balance.
• Wrong speaker, member, party, vote, date, or source attribution.
• Broken source link or unsupported provenance.
• Official/product boundary confusion.
• Copyright or licence misuse.
• Defamation, unfair allegation, or legal-sensitive framing.
• Unsafe amplification of harmful parliamentary statements without context.
• Ranking, recommendation, inclusion, grouping, or curation skew.
• Inaccessible correction, redress, or appeal route.
• Future user-conduct harm if interactive features launch.
• Child or vulnerable-user risk if future interactive or search-like features launch.

Severity Model

| Severity | Description | Typical routing | | --- | --- | --- | | Low | Typo, minor copy issue, non-material metadata issue | System triage or editorial backlog | | Medium | Incorrect source link, unclear AI label, missing context that may mislead some users | Fact-check or editorial review | | High | Wrong attribution, materially misleading summary, public complaint from an affected person | Editorial review plus escalation consideration | | Critical | Possible legal harm, unsafe content exposure, repeated system failure, many affected items, or regulatory concern | Safety lead, legal-sensitive escalation, and incident review |

Queue Mapping

Risks should route to AI/system triage, fact-check/editorial review, escalation review, legal-sensitive review, transparency-log review, or future moderation review depending on severity and content class.

Regulatory Discovery

Current assumptions should be reviewed before launching user-to-user, search-like, comment, debate, opinion, reaction, sharing, or personalised ranking features. Official UK online-safety references should be treated as review inputs, not as a public legal conclusion.