Do not rely blindly on AI-generated content, expert says

SHAH ALAM - Artificial intelligence (AI) is revolutionising content creation, but without proper human review, it can also produce dangerously inaccurate outputs, an expert warns.

AI trainer and consultant Cletus de Rozario has cautioned organisations and the public against placing blind trust in AI-generated content, especially when dealing with culturally or nationally sensitive material such as the Jalur Gemilang, Malaysia’s national flag.

“Any content intended for public consumption, particularly if generated by AI, must be thoroughly reviewed before publication. While AI is a powerful tool, it is not immune to mistakes,” he told Sinar Daily.

Cletus highlighted a common issue known as ‘AI hallucination’, where the system generates believable but factually incorrect information.

These hallucinations can occur in both text and visuals, the latter sometimes called ‘image hallucinations.’

He explained that such errors can stem from a range of causes, including vague prompts, limited data on niche topics or the AI’s inability to grasp nuanced context.

As an example, he pointed to the potential risks in generating a symbol as meaningful as the Jalur Gemilang.

Cletus emphasised that most public AI models do not access real-time or verified data but rely instead on patterns learned from large training datasets.

This limitation increases the risk of errors in outputs.

“This means that while it is technically possible for AI to generate a flawless version of the Jalur Gemilang, success depends on three key factors: the quality and accuracy of the training data, the clarity and specificity of the prompt and most importantly, human review.

“A clear, context-rich prompt is like asking a smart person a focused question; it improves the likelihood of receiving a relevant and accurate answer. That is why prompt engineering matters.

“Many users, accustomed to everyday language, may unintentionally give vague or incomplete prompts that lead to hallucinated content. When the input is unclear, the AI tries to ‘fill in the gaps,’ which often results in convincing but incorrect responses,” he added.

Cletus warned that the recent misuse or errors seen in AI-generated content may very well stem from inexperience or a lack of review.

While it is tempting to use AI as a creative co-pilot, he said proper training is essential.

“We need to equip users with prompt engineering skills and integrate AI literacy into organisational practices. Training should not be optional, it is essential.”

“Just as important is the culture of accountability. Organisations must make it standard policy to verify AI-generated content, especially when dealing with sensitive elements like national symbols.

“Relying blindly on AI without checks is risky, particularly in multicultural and multireligious societies like Malaysia, where a mistake involving a cultural or national icon can trigger significant backlash.

“This is why we need strong governance and ethical frameworks,” he highlighted.

He also argued against blanket bans on AI, such as prohibiting it from generating national symbols, suggesting such measures may be ineffective in the long run.

Rather than adding new restrictions each time an error occurs, he believed the priority should be on setting clear guidelines, safeguards and encouraging responsible use.

“AI ethics and governance are still evolving in Malaysia and around the world, but there is encouraging progress. More frameworks, codes of conduct and regulatory policies are being developed to support the responsible use of AI.

“Ultimately, a hybrid approach is the most sustainable way forward. Humans and AI work together, with each complementing the other.

“We may not eliminate errors completely, but we can significantly reduce the risks through proper review, training and thoughtful design of AI workflows,” he added.

He stressed the need for verified tools, trusted data and expert review panels, especially in media and education sectors.

By keeping detailed records of what is generated and how it's used, transparency and accountability will be added to the process.

“In short, we should use AI carefully, with human judgment at the core, ensuring that the technology aligns with Malaysia’s values and respects its diverse cultural identity,” he said.

Between April 15 and April 29, several incidents involving inaccurate representations of the Jalur Gemilang, Malaysia's national flag, triggered public outrage.

Between April 15 and April 29, multiple incidents involving inaccurate depictions of the Jalur Gemilang; Malaysia’s national flag, sparked public outrage, formal apologies and even political commentary.

The first controversy erupted on April 15, when one of Malaysia’s oldest Chinese-language newspapers published a front-page image of the Malaysian flag without the crescent moon, an Islamic symbol central to the flag’s identity.

The image was meant to celebrate Chinese President Xi Jinping’s state visit but instead triggered backlash for omitting a religiously significant emblem.

The newspaper quickly apologised and corrected the digital version, blaming the incident on a design error, with critics pointing to over-reliance on AI and lack of editorial review.

On April 20, a company participating in a baby expo in Kuala Lumpur displayed a promotional video featuring a similarly incomplete Jalur Gemilang, again missing the crescent.

The incident received at least five police reports and prompted an apology from the company, which acknowledged the gravity of the mistake and vacated its booth from the event.

The issue resurfaced on April 24, when the Education Ministry (MOE) admitted to a flag design error in its Sijil Pelajaran Malaysia (SPM) examination results analysis report, where a flag with two stars and only eight stripes appeared in an official visual.

The ministry issued a formal apology, launched an internal investigation and ordered the recall of the faulty report.

By April 29, former prime minister Tun Dr Mahathir Mohamad weighed in, claiming that the repeated errors were not coincidental.

He alleged the removal of the crescent moon was intentional and possibly aimed at aligning Malaysia with communist symbolism, referencing flags of communist nations that use only stars.

The controversies have stirred national debate over the sanctity of national symbols, accountability in public institutions and the implications of such oversights.

Whether accidental or as some suggested, deliberate.