The debunking method here is the practical template worth noting: AFP treated OpenAI's image-verification signal as one input among several, cross-checking it against official photo releases and garment-level detail on the queen's outfit, rather than trusting the AI-detection tool's verdict alone.
What happened
An official from Thailand's foreign ministry told AFP on July 2, 2026 that an image circulating online, purporting to show French President Emmanuel Macron kneeling before King Maha Vajiralongkorn while presenting a royal decoration during a state visit, was AI-generated, stating "there is no evidence of any image showing Emmanuel Macron kneeling to present a royal decoration." AFP's Factcheck service ran the image through OpenAI's image verification tool, which it reported determined the picture was "generated with OpenAI tools." AFP separately found no official photographs from Thailand's foreign ministry showing Macron kneeling at the decoration ceremony or state dinner, and identified outfit differences when comparing the woman in the fabricated image to AFP's own photographs of Queen Suthida. One Thai-language Facebook post of the image drew more than 40,000 likes and 2,000 shares by the following day and was amplified by a page with over 2 million followers, according to AFP reporting syndicated by Arab News, France 24, the South China Morning Post, and other outlets.
For practitioners
AFP's workflow is a reusable template for triaging viral visual claims: run an automated provenance or AI-detection check and log the raw output, cross-reference against primary or official sources for temporal or contextual mismatches (here, the absence of any official kneeling photo and a garment mismatch), and preserve the original post with engagement metrics before it is taken down. None of these three steps alone was conclusive; the AI-detection result, the missing official photo, and the outfit mismatch together built AFP's case, which is the more durable lesson for anyone building automated moderation or verification pipelines than the specific tool used.
What to watch
Watch for whether AFP or OpenAI disclose more detail on how the image-verification tool reached its determination (confidence scores, detected provenance metadata such as C2PA credentials), since factcheckers currently report binary tool outputs without much transparency into the underlying signal. Also watch for repeat incidents involving state visits and royal or diplomatic protocol, a recurring target for viral fabricated imagery given its high shareability.
Key Points
- 1Thailand's foreign ministry told AFP a viral image of Macron kneeling before King Vajiralongkorn during a state visit was AI-generated.
- 2AFP verified the fake using three checks together: OpenAI's image-verification tool, missing official photos, and an outfit mismatch versus real queen photos.
- 3One Facebook post of the fake image drew over 40,000 likes before debunking, illustrating how fast synthetic diplomatic imagery can spread.
Scoring Rationale
A well-corroborated (7 outlets syndicating the same AFP wire report), single-incident case of AI-generated diplomatic misinformation with a genuinely useful multi-signal verification template for practitioners building moderation or provenance pipelines. Held in the solid range since it is one debunked incident rather than a systemic shift in detection technology or policy.
Sources
Public references used for this report.
View 4 more sources
- 04Thailand denies viral claim Macron knelt before kingthehindu.com
- 05Thailand denies viral claim Macron knelt before kingnewsofbahrain.com
- 06Thailand denies viral claim Macron knelt before kingfreemalaysiatoday.com
- 07Thailand denies AI-generated image of Macron kneeling before king during state visit to France, calling it fakethesun.my
Practice interview problems based on real data
1,625 SQL & Python problems across 15 industry datasets — the exact type of data you work with.
Try 250 free problems
