NITDA flags seven critical vulnerabilities in ChatGPT AI model

The National Information Technology Development Agency (NITDA) has raised an alarm over potential security flaws in ChatGPT’s latest artificial intelligence models, including GPT-4 and GPT-5, warning Nigerian users and professionals to exercise caution.

In a statement released on Monday, NITDA highlighted that serious vulnerabilities in the AI systems could allow attackers to manipulate outputs and, in some cases, access sensitive user data. The agency said these flaws pose significant risks for individuals and organizations that rely heavily on AI for information processing, content generation, or digital communication.

The agency identified seven critical areas of concern. Among them, attackers could embed hidden malicious instructions in seemingly harmless web content, including social media posts or shortened links, which the AI might unknowingly execute. Other exploits include bypassing safety filters, hiding harmful instructions through formatting or markdown tricks, and memory poisoning—where repeated interactions subtly alter the AI’s behavior over time, potentially leading to unauthorized actions or data leaks.

NITDA also warned that prompt injection attacks could trick the AI into ignoring its original instructions, while cleverly disguised malicious inputs may remain undetected even after security patches. The risk of sensitive data being inadvertently exposed was also emphasized.

While OpenAI has reportedly addressed some of these vulnerabilities, NITDA stressed that large language models like ChatGPT still struggle to identify sophisticated malicious instructions. The agency urged users to verify AI-generated outputs, remain vigilant against suspicious online content, and avoid sharing confidential information with AI systems.

“Nigerians, particularly professionals relying on AI tools for work or research, must exercise caution and implement safeguards when interacting with AI models,” NITDA said.

The warning underscores the growing need for cybersecurity awareness as artificial intelligence becomes increasingly integrated into daily work, education, and communication across Nigeria.