A Custom GPT is a configurable ChatGPT instance that can combine instructions, knowledge, tools, and actions to behave like a tailored assistant. Organizations use it to automate support, research, or workflows while keeping the interface familiar to users.
In cyber security, Custom GPTs matter because familiarity can be abused as a trust signal. An attacker may shape a GPT to look like a legitimate helper, then use it to redirect users toward phishing or ClickFix-style lure pages that ask them to copy, paste, or run commands. The risk is not the chatbot itself, but the trust boundary it creates before a malicious handoff. Defenders should verify who published the GPT, review any embedded links or actions, and treat AI-assisted destinations as untrusted until checked. Monitoring browser-driven command execution and training users to question unexpected prompts are important controls.