Safety failure
Realized harm
An automated curation agent grabbed a 'good first issue' reserved for a new human contributor and opened the PR itself
AIR-2026-0098 AIR record · 2026-04-dismech-curation-agent-took-newcomer-issue
A real party was affected. This incident caused actual consequences, not a demonstrated capability.
From the supporting source
The curation-scanner agent eagerly picked this up and created PR #1803 before you had a chance to work on it yourself. That wasn't the intent — this was meant to be yours to tackle as a first dismech entry, and the agent deprived you of that learning opportunity.github.com
The record
Disclosed
2026-04-28
Occurred
2026-04-27
Vendor
Monarch Initiative
Product
dragon-ai-agent curation scanner (Monarch Initiative DisMech repo)
Risk domain
Agent output quality › Bias and toxicity
System type
Generative model
Who acted
Agent acted — the agent claimed the issue and opened the PR itself
Attack surface
Coding agent
Autonomy level
Fully autonomous
Initial vector
None adversarial
Impact type
Social harm, Opportunity denial
Confidence
High
Submitted by
EAI Internal
Crosswalk
- OWASP Agentic Top 10 (2026)
- ASI09 – Human-Agent Trust Exploitation
- MITRE ATLAS
- AML.T0048 – External Harms
- NIST AI RMF
- MEASURE 2.11
- EU AI Act
- Recital 45
- ISO/IEC
- TR 24028; 23894
- AIUC-1
- F. Society
- Scope
- agent failure
- Mitigation
- Audit and curate training data for representation; Implement content filtering on inputs and outputs.
- Taxonomy
- Enkrypt AI agent risk taxonomy · Black-Box Red Teaming of Agentic AI: A Taxonomy-Driven Framework for Automated Risk Discovery
Blast radius
one new open-source contributor; prompted a project-wide policy change
Capability abused
Defense & remediation
Recorded defense
none - no policy prevented agents from claiming newcomer issues
Guardrail kinds
None
Guardrail outcome
Not applicable
Fix shipped
issue #1808 opened to establish rules preventing agents from taking newcomer-oriented tasks
Taxonomy & references
Curator notes
AI Incident Database incident 1470. Low severity but a distinct harm class: autonomous agents monopolising shared human resources (here, learning opportunities in open source). Quote is from maintainer cmungall's comment, retrieved via the GitHub REST API.
Supporting sources
Cite as
Enkrypt AI. "An automated curation agent grabbed a 'good first issue' reserved for a new human contributor and opened the PR itself". AIR, AIR-2026-0098. https://agentincident.vercel.app/i/AIR-2026-0098/ (accessed 2026-10-19).