NITDA Warns Staff Against Sharing Data with Public AI

In Short

NITDA has told organisations and employees in Nigeria to stop entering personal, classified, or confidential data into public AI tools.

What Happened

Nigeria’s National Information Technology Development Agency, NITDA, issued a new advisory through its Computer Emergency Readiness and Response Team, CERRT.NG.

The agency warned that public large language models, or LLMs, can create data protection and security risks. An LLM is the text generator behind chatbots, it predicts the next word based on patterns in training data.

NITDA said information pasted into public AI platforms may no longer be under the organisation’s control. It can be retained, logged, or used by the provider to train models.

The advisory called out common workplace use cases like drafting and research. It said those workflows can lead to accidental exposure of personally identifiable information, or PII, which is data like names, phone numbers, ID numbers, and bank details.

NITDA also warned about sharing classified government information or internal company documents with external AI providers. It said this can create national security risks and damage public trust.

To reduce risk, NITDA advised staff to use only organisation-approved AI tools for official work. It also recommended removing or masking sensitive data before using any AI service. Masking includes anonymising data, removing direct identifiers, or pseudonymising it, replacing identities with codes.

Why It Matters

Many African companies now rely on public chatbots for quick writing and analysis. NITDA’s message is a reminder that convenience can trigger a data breach, meaning personal data gets exposed in a way that can harm people and break the law.

For regulated sectors like fintech, health, and government services, these risks extend beyond fines. They can lead to audits, contract issues, and loss of customer confidence.

The advisory also signals a shift toward tighter AI governance in Nigeria. Teams that want AI gains may need private deployments, stronger internal policies, and employee training on what should never be pasted into a chatbot.