Threat Actors Abuse Generative AI for Scams, Influence Operations, and ClickFix Social Engineering

OpenAI reported multiple cases of malicious use of ChatGPT and related API access spanning financially motivated fraud and government-aligned information operations, including romance scams, fake legal services, coordinated influence campaigns, and a state-linked harassment effort. In one detailed case (“Operation Date Bait”), actors used ChatGPT accounts and an API customer to generate ad copy for a fake dating service, drive victims to Telegram, and run semi-automated “missions” that required escalating payments; OpenAI said it banned accounts involved and noted the group’s claimed revenues could not be independently verified. OpenAI also described model use for translation, persona development, and internal coordination (e.g., generating status reports and assigning targets projected payout values), and cited evidence such as repeated text patterns posted by multiple X accounts.

Google’s Threat Intelligence Group (GTIG) separately assessed that nation-state threat actors are operationalizing LLMs (including Gemini and other tools) for reconnaissance, victim targeting, and generating culturally fluent, multi-turn “rapport-building” phishing lures. GTIG also observed abuse of generative AI services’ public sharing features to host deceptive content used in ClickFix campaigns, where victims are tricked into copying and pasting malicious commands into a terminal; the activity was first observed in early December 2025. The reporting highlights a converging trend: adversaries are using LLMs to scale and localize social engineering while shifting delivery mechanisms toward trusted AI platforms and shared content features to increase credibility and reach.