Intellectual Worship in the Era of AI
On 1 June 2026, coinciding with the Birth of Pancasila, the National Research and Innovation Agency (BRIN) uploaded a commemorative poster via its official social media accounts. However, rather than being a source of pride, the post triggered a wave of public criticism. Netizens discovered several discrepancies in the Garuda Pancultila image used, particularly regarding the number of feathers on the wings, tail, and neck, which did not comply with the official provisions of Law Number 24 of 2009.
BRIN subsequently issued an apology and replaced the post with a corrected version. This case is noteworthy not only because of the design error but also due to allegations that the image was created using generative AI. The visual characteristics displayed resemble common weaknesses found in AI-generated outputs: appearing convincing at a glance but erroneous in critical details. Regardless of whether these allegations are true, this incident serves as a reminder that technology capable of producing content rapidly is not necessarily capable of understanding the underlying meaning of the symbols it generates.
In the realm of artificial intelligence, this condition is known as ‘AI hallucination’—where a system produces outputs that appear correct and convincing but are actually factually incorrect. In the case of the Garuda Pancasila image, every feather holds historical significance representing Indonesia’s independence date, 17 August 1945. When these details are incorrect, the symbolic meaning within the image is lost.
This follows a recent case involving alleged research forgery at the ISPPD 2026 conference in Copenhagen, where AI was suspected of being used to generate professional-looking abstracts and scientific posters containing non-existent data. Both the BRIN and Copenhagen cases highlight the same issue: delegating outputs to AI without adequate verification processes. While the cases differ, both lead to the same consequence: the erosion of public trust.
In academic environments, AI hallucinations often manifest in more dangerous and harder-to-detect forms, such as fabricated references or citations. A study published in The Lancet in May 2026, led by researchers from Columbia University, audited approximately 2.5 million biomedical articles published over the last three years. The results showed a worrying trend. In 2023, about one in every 2,828 articles contained at least one fabricated reference. By 2025, this figure rose to one in 458 articles. In the first seven weeks of 2026, the ratio reached one in 277 articles. In just three years, the presence of fake references in scientific publications has increased more than twelvefold.
AI-generated references are undeniably convincing. Author names appear plausible, article titles sound scientific, journals appear reputable, and even the displayed DOI numbers look valid at first glance. However, upon closer inspection, these references do not exist. For students using them in theses or dissertations without verification, the consequences can be severe: delayed defences, rejected works, and questioned academic integrity. For researchers and lecturers, the impact is even greater as it affects professional reputations built over years.
Using AI as a writing aid is natural and increasingly necessary. However, every piece of information, citation, data, or image generated by AI must be independently verified. One must check if references truly exist, ensure authors and journals are real, and verify DOIs and source links. Just as the Garuda Pancasila image should have been checked before publication, every AI output in academic work must undergo critical human assessment.
AI is an extraordinary tool, but it lacks responsibility; that responsibility lies with us. Errors in the Garuda Pancasila image can be rectified through apologies and re-uploads, but fake references that enter scientific literature are much harder to purge because they can be cited by subsequent research. In an era where AI is becoming increasingly adept at generating information, the human ability to examine, verify, and take responsibility for information becomes even more vital. AI can generate answers, but it cannot bear the responsibility for their truth. That responsibility remains in human hands.
Therefore, the process of fact-checking, tracing sources, testing data, and ensuring the truth of information is not merely a part of scientific methodology, but a form of ‘intellectual worship’. Through verification, we ensure that knowledge remains grounded in truth rather than mere appearance. In the era of AI, this is perhaps one of the most important scientific trusts to uphold.