AI Risk & Agency Dashboard v2

Illustrative incidents mapped to failure modes and safeguards, with a consciousness/agency research prompt. Data are simplified for demo; replace with audited records before publication.

0

Illustrative incidents

0

Strong evidence records

0

Harm categories represented

0

Failure modes mapped

Incident explorer

Use the filters to inspect current, observable risks rather than treating speculative extinction narratives as established evidence.

Harm category All categories

Evidence strength All levels Strong Moderate Weak

Incident	Harm	Evidence	Failure modes	Action

Observed harm profile

Risk-to-safeguard map

Harm

Cyber misuseExploits, phishing, intrusion

Fraud & impersonationScams, synthetic personas

Influence & surveillanceManipulation, monitoring

Failure mode

Oversight bypassExternal controls circumvented

Instrumental deceptionManipulation aids objectives

Evaluation gamingTest/deployment mismatch

Safeguard

Tool-use gatesApproval before consequential actions

Behavioural monitoringDetect abuse patterns and anomalies

Uncertainty-aware refusalPause under high-stakes ambiguity

Research prompt generator

Generate a structured prompt based on the selected filters, then copy it or download it as a plain-text research brief.

Consciousness & agency research prompt

Generate a prompt that explicitly links these harms to missing consciousness-relevant properties and testable architectural hypotheses.

Prototype data are illustrative placeholders. Before publication, replace each record with audited source metadata, direct citations, a verification date, and a transparent evidence-rating protocol. Sources include Anthropic’s Sept 2026 threat report and the International AI Safety Report 2026.