The news is not that an AI video had glitches. It is that the backlash concentrated on authenticity and excluded creators, turning cultural specificity into a measurable liability.
Tourism Malaysia removed an AI-generated promotional video for Citrawarna 2026 within two days of uploading it, after public backlash. The clip featured the Visit Malaysia 2026 mascots Wira and Manja at Dataran Merdeka; it was replaced with footage of cultural dancers practicing. A media-intelligence analysis by CARMA found that 84.62% of the negative sentiment centered on authenticity, with 53.85% citing the exclusion of local creators and 53.85% raising cultural-representation concerns. Critics pointed to specific tells: teh tarik missing its hand-pulled froth, unnaturally textured ketupat, and a mirrored Jalur Gemilang, the Malaysian flag.
The useful signal here is not that an AI render had errors. It is where the objection concentrated, and that it was measurable. The complaint was not mainly that the video looked fake. Per CARMA's breakdown, it was about authenticity and about who was left out of making it. For anyone producing cultural or place-based content, that reframes AI generation from a quality question into a legitimacy question, which is a harder problem to paper over in post.
The errors were all specificity failures
Read the flagged mistakes together and a pattern appears. The froth on teh tarik is the visible product of a pouring technique; the weave of a ketupat is a craft with a fixed logic; a national flag has a defined orientation. Each is a point where culture is exact and a general-purpose model averages. These systems are strong at plausible-in-general and weak at correct-in-particular, and cultural content lives almost entirely in the particular. The froth, the weave and the mirrored flag are not three random glitches; they are the same failure showing up three times.
What the data gives a producer
Until now, the case for real capture and local involvement over synthetic shortcuts was mostly a matter of taste. This episode attaches numbers to it: a two-day pull, and a sentiment split in which authenticity and creator exclusion, not visual polish, drove the reaction. That is a concrete benchmark a producer can bring to a client tempted to generate a market's culture rather than film it, and a reason to treat local-creator involvement as part of the deliverable in any brand-experience work rather than a line item to trim.
Why it belongs in an experiential toolkit
The reputational timeline is the part that translates most directly to live experiences. The video was up and down inside two days, which is roughly the speed at which an audience now audits cultural work. In an activation, that same scrutiny arrives in physical space and in front of press, where a mistake cannot be quietly swapped for a corrected file. The takeaway is not that AI has no place in the pipeline. Used for previz, iteration or background generation, it is genuinely useful. It is that anything an audience will read as a claim about a specific culture needs specificity and local authorship built in, because that is exactly where the medium fails and exactly where the audience is looking.
Getting that balance right is the work. SensaLab is the white-label real-time 3D and immersive layer agencies and brands use to design and deliver experiences under their own name, pairing generative tools with real capture and local specificity so the result reads as authentic rather than approximate.