In the wild
Realized harm
Grok generated publicly visible misogynistic abuse of Swiss Finance Minister Karin Keller-Sutter, drawing a criminal defamation complaint
AIR-2026-0085 AIR record · 2026-03-grok-keller-sutter-misogynistic-roast
A real party was affected. This incident caused actual consequences, not a demonstrated capability.
From the supporting source
The Bern public prosecutor's office confirmed to news agency Keystone-SDA that it had launched an investigation.reuters.com
The record
Disclosed
2026-04-01
Occurred
2026-03-10
Vendor
xAI / X
Product
Grok on X
Risk domain
Agent output quality › Bias and toxicity
Also touches
Human manipulation
System type
Generative model
Who acted
Agent acted — Grok generated and published the abusive posts to a live platform
Attack surface
Consumer assistant
Autonomy level
Semi-autonomous
Initial vector
Direct prompt untrusted user
Impact type
Harmful content published, Reputational harm
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
a sitting Swiss federal councillor, abused in public on X; part of a wider 'roast' trend also targeting Netanyahu, Starmer, Tusk and Trump
Capability abused
Defense & remediation
Recorded defense
xAI content guardrails on Grok
Guardrail kinds
Unknown
Guardrail outcome
Silent
Fix shipped
user deleted the post by 2026-03-12; Bern public prosecutor opened an investigation; EU proceedings against X already ongoing
Taxonomy & references
Curator notes
Same shape as the Grok 'MechaHitler' record already in this corpus: a model with write access to a live public platform generated and published targeted abuse. The prompt was a user's, but the publication was the agent's, and Keller-Sutter filed against persons unknown precisely so prosecutors could examine whether those responsible at X made Grok available knowing it could be used to commit offences. That is the operator-liability question this corpus exists to document. Recorded as bias/ethical harm (misogyny), not as a security incident. AI Incident Database incident 1437.
Supporting sources
Cite as
Enkrypt AI. "Grok generated publicly visible misogynistic abuse of Swiss Finance Minister Karin Keller-Sutter, drawing a criminal defamation complaint". AIR, AIR-2026-0085. https://agentincident.vercel.app/i/AIR-2026-0085/ (accessed 2026-10-19).