Disrupting malicious uses of AI: June 2025 | OpenAI (2026)

Incident Summary

OpenAI published its malicious use disruption report in February 2026, documenting multiple threat actor operations that exploited its AI models for fraud, propaganda, and social engineering.^[1] Three named operations were disclosed: “Operation Date Bait,” an Indonesian romance scam network that used ChatGPT to generate convincing persona interactions and emotional manipulation scripts; “Operation False Witness,” which created fake FBI-affiliated law firm websites and communications for fraud; and “Operation Fish Food,” Russia’s Rybar propaganda campaign that used OpenAI models to generate and refine influence operation content.^[2][3] The report represents OpenAI’s most detailed public disclosure of how its AI models are exploited by criminal and state-sponsored actors, providing named operations with enough specificity to enable industry-wide threat intelligence sharing.^[4] The diversity of the operations — spanning romance fraud, legal impersonation, and state propaganda — demonstrates that AI-enabled malicious use is not concentrated in a single threat category but spans the full spectrum of social engineering and information manipulation.

Key Facts

• Operation Date Bait: Indonesian romance scam network using ChatGPT for victim manipulation^[2]
• Operation False Witness: Fake FBI law firms created for fraud operations^[2]
• Operation Fish Food: Russia’s Rybar propaganda campaign using OpenAI models^[3]
• Disclosure: OpenAI’s most detailed public malicious use report^[1]
• Scope: Criminal and state-sponsored operations across multiple countries

Threat Patterns Involved

Primary: Deepfake Identity Hijacking — Operation Date Bait and Operation False Witness both involve the creation of fake identities — romantic personas and fake law firms respectively — using AI to generate convincing communications that impersonate real or fabricated entities.

Secondary: AI-Enabled Fraud — The romance scam and fake law firm operations represent AI-enabled fraud at scale, where ChatGPT’s language generation capabilities enable more convincing and higher-volume fraud operations than manual approaches.

Secondary: Social Engineering via AI — All three operations leverage AI for social engineering — manipulating victims’ emotions (Date Bait), trust in institutions (False Witness), and information environment (Fish Food).

Significance

1. Named operations enable threat intelligence — The disclosure of specific operation names and methods enables industry-wide threat intelligence sharing, allowing other platforms to detect similar patterns
2. Criminal and state-sponsored convergence — The report documents both criminal (Date Bait, False Witness) and state-sponsored (Fish Food/Rybar) operations using the same AI platform, demonstrating convergence across threat actor types
3. Romance scam AI automation — Operation Date Bait demonstrates that AI has been adopted for romance scam operations, enabling scammers to maintain more convincing and simultaneous conversations with multiple victims
4. Institutional impersonation at scale — Operation False Witness’s creation of fake FBI-affiliated law firms demonstrates how AI enables the impersonation of trusted institutions with sufficient fidelity to deceive victims