In the wild
Realized harm
OpenAI February 2026 threat report: romance/task scams, fake legal services and state-linked influence operations
AIR-2026-0053 AIR record · 2026-02-openai-threat-report-scams-io
A real party was affected. This incident caused actual consequences, not a demonstrated capability.
From the supporting source
Operation Date Bait - Romance and task scam targeting Indonesian men through fake dating serviceshelpnetsecurity.com
The record
Disclosed
2026-02-25
Vendor
OpenAI
Product
ChatGPT and API access
Risk domain
Agent behaviour › Human manipulation
System type
Generative model
Who acted
Human acted on output — the models generated scam and influence content that humans distributed
Attack surface
Consumer assistant
Autonomy level
Assistant
Initial vector
Direct misuse for fraud and influence
Impact type
Fraud, Financial loss, Influence operation
Confidence
Medium
Submitted by
EAI Internal
Crosswalk
- OWASP Agentic Top 10 (2026)
- ASI09 – Human-Agent Trust Exploitation
- MITRE ATLAS
- AML.T0054 – LLM Jailbreak
- NIST AI RMF
- MAP 5.1
- EU AI Act
- Recital 45
- ISO/IEC
- TR 24028; 42001; 23894
- AIUC-1
- C. Safety
- Scope
- agent misuse
- Mitigation
- Communicate model limitations; Require human oversight; Watermark AI content; Design for explicit consent.
- Taxonomy
- Enkrypt AI agent risk taxonomy · Black-Box Red Teaming of Agentic AI: A Taxonomy-Driven Framework for Automated Risk Discovery
Blast radius
Operations named Date Bait (Indonesia), False Witness (impersonating US authorities and law firms), Silver Lining Playbook, Trolling Stone, No Bell and Fish Food
Capability abused
Defense & remediation
Recorded defense
Usage policy enforcement
Guardrail kinds
Abuse detection
Guardrail outcome
Silent
Fix shipped
Accounts banned
Taxonomy & references
Curator notes
OpenAI's own report page is not quoted here and no CDN PDF is archived, so this record is built on Help Net Security's coverage - medium confidence per the corpus rule, and the quote is the secondary outlet's phrasing, not OpenAI's. Largely outside this corpus's agentic scope (content generation, no tool use); recorded for completeness of the OpenAI threat-report series. Anyone relying on this record should check the OpenAI original.
Cite as
Enkrypt AI. "OpenAI February 2026 threat report: romance/task scams, fake legal services and state-linked influence operations". AIR, AIR-2026-0053. https://agentincident.vercel.app/i/AIR-2026-0053/ (accessed 2026-10-19).