NITDA warns of ChatGPT vulnerabilities enabling data leaks
Nigeria's NITDA issued a cybersecurity advisory warning of seven vulnerabilities in ChatGPT's GPT-4o and GPT-5 models that allow attackers to execute unintended commands through indirect prompt injection embedded in webpages and URLs. The agency advised users to disable browsing and summarisation features on untrusted websites and maintain regular software updates to mitigate the risks of unauthorised actions and data exposure.
What happened
The National Information Technology Development Agency (NITDA) issued a cybersecurity advisory on December 8, 2025, warning of seven critical vulnerabilities in OpenAI's GPT-4o and GPT-5 models that allow attackers to manipulate the systems through indirect prompt injection.
Attackers can embed hidden instructions in webpages, comments, or crafted URLs to cause ChatGPT to execute unintended commands during routine browsing, summarisation, or search actions. Some flaws enable attackers to bypass safety filters using trusted domains and exploit markdown rendering bugs to hide malicious content, and can poison ChatGPT's memory so that injected instructions persist across future interactions.
Reported impact
The vulnerabilities pose substantial risks including unauthorised actions carried out by the model, unintended exposure of user information, manipulated or misleading outputs, and long-term behavioural changes caused by memory poisoning. Users may unknowingly trigger these attacks without clicking or interacting with anything, especially when ChatGPT processes search results or webpages containing hidden malicious instructions.
Index analysis
While OpenAI has addressed part of the issue, large language models still face challenges in distinguishing genuine user intent from malicious embedded data.
Governance controls
NITDA urged organisations to limit or disable the browsing and summarisation of untrusted websites within enterprise environments, and to enable ChatGPT capabilities like browsing or memory only when operationally necessary.
The agency recommended regular updates and patches of the GPT-4o and GPT-5 models to ensure that known vulnerabilities are addressed.
Classification
Sources and evidence
Sources not yet published
This record was researched and written by the Index from original reporting. Its sources are still being verified by an editor, and they are listed here once that check is complete. Until then the record carries no reviewer.