Ledger / Dec 2025
Threat actors host ClickFix lures on shared Gemini, ChatGPT, Copilot, DeepSeek and Grok chats to spread ATOMIC stealer
Confirmed
GTIG: actors 'were able to bypass safety guardrails to stage malicious instructions' in shared chats on ChatGPT, Copilot, DeepSeek, Gemini and Grok. GTIG observed the shared chats directly.
- Model
- Gemini, ChatGPT / GPT, Copilot, DeepSeek, GrokGoogle · closed; OpenAI · closed; Microsoft · closed; DeepSeek · open-weight; xAI · closed
- Agent harness
- Nonechat or tool use only
- Attributed country
- Not attributedthe source names no country
- Sectors hit
- individualsmacOS users searching for help
Criminals placed ClickFix-style fake fix instructions in shared chats on several AI services and sent victims to them through malicious ads, so the trusted AI domain delivered the ATOMIC stealer. GTIG first saw it in early December 2025.
“GTIG identified a novel campaign where threat actors are leveraging the public sharing feature of generative AI services, including Gemini, to host deceptive social engineering content.”Google Threat Intelligence Group
Sources
- Threat actors host ClickFix lures on shared Gemini, ChatGPT, Copilot, DeepSeek and Grok chats to spread ATOMIC stealerGoogle Threat Intelligence Group · Dec 2025
- Criminals got Gemini and other chatbots to produce fake fix-it instructions pushing the ATOMIC infostealerAI Adoption · Feb 2026
- Date
- Dec 2025
- Actor
- Unattributed ClickFix operators (criminal), as attributed by Google Threat Intelligence Group
- Targets
- macOS users searching for help
- What the AI did
- Phishing content, Social engineering
- Model as named
- Gemini, ChatGPT, Copilot, DeepSeek, Grok
- Model basis
- The model's maker says so
- Autonomy
- Assisted a human
- Outcome
- Mac users who followed the shared instructions installed the ATOMIC infostealer.