The National Information Technology Development Agency (NITDA) has raised an urgent cybersecurity warning concerning weaknesses identified in OpenAI’s GPT-4 and GPT-5 artificial intelligence models. The agency cautioned that these vulnerabilities could be exploited by cybercriminals to influence AI responses and potentially obtain unauthorised access to confidential user information.
In its advisory, NITDA explained that a detailed technical review uncovered several security loopholes that pose risks to individuals and organisations using the technology. One of the most critical threats is known as indirect prompt injection, a method that enables attackers to embed harmful instructions within ordinary online content such as social media posts, blogs, or shortened web links. When an AI system is asked to analyse or summarise such material, it may unknowingly execute the concealed commands.
According to the agency, this form of manipulation can result in serious consequences, including data exfiltration—where sensitive information is secretly transmitted to external servers. Another danger is memory poisoning, an advanced tactic that gradually alters the AI’s stored information, leading to long-term bias or distorted outputs. NITDA also pointed to flaws in markdown rendering that could be abused to bypass built-in safety mechanisms, forcing the model to generate prohibited or harmful content.
NITDA Advises Vigilance as AI Risks Persist
Although OpenAI has reportedly acknowledged some of these shortcomings and introduced preliminary fixes, the agency stressed that large language models still struggle to reliably differentiate between genuine user instructions and cleverly disguised malicious input. It warned that users could be compromised simply by allowing the AI to process a poisoned search result, even without taking any further action.
As part of its recommendations, NITDA urged Nigerian professionals and businesses to exercise greater caution when relying on AI tools. Users are advised to double-check AI-generated information, particularly when it relates to technical, financial, or personal data. Organisations were also encouraged to remain vigilant for suspicious online material and to monitor AI systems for unusual behaviour that might indicate a breach.
The agency reiterated its commitment to safeguarding Nigeria’s digital ecosystem, stating that it will continue to track developments in artificial intelligence and issue guidance to protect citizens and institutions from emerging cyber threats.
Stay informed on critical cybersecurity and technology developments in Nigeria. Visit LouisaOlaniyi.com.ng for more expert analysis, verified tech news, and digital policy updates.
