Kalshi’s AI ad turned a real Asian American creator white

What happened to the video

The original video documented Choi’s move into a new apartment in New York. He published it in September last year. Kalshi took that footage and changed the person on screen before making him advertise the company’s product.

Choi discovered the ad only hours after Stanford University was accused of using AI to replace a student in a photograph with a fictional Black woman. Choi himself had used advertising to promote Anthropic’s Claude assistant, making the contrast hard to miss: one creator was paid to appear in an ad, while his own footage was repurposed without permission for another.

NPR correspondent Bobby Allin confirmed that the Kalshi advertisement existed. A company representative told him the video was old and that Kalshi no longer uses AI-generated advertising. The company also said it was reviewing its relationship with the agency that produced the video.

The unanswered questions are straightforward:

The advertising problem behind it

Allin first noticed a similar pattern in July last year. TikTok was filled with videos that copied user-created clips using deepfake technology, with some later reaching tens of millions of views.

Most of those videos appeared to come from scammers rather than major companies seeking advertising material. Digital forensics expert Hany Farid of the University of California told Allin that current AI tools make this kind of manipulation easy to produce, allowing it to pass through content-moderation systems without attracting attention.

Kalshi has been experimenting with generative AI in marketing for years. The company previously showed an AI-generated advertisement for the NBA Finals and said it cost only $2,000 to produce. Whether that saving justified the result was already debatable.

The company later moved in the opposite direction as criticism of AI-generated content grew:

Keaton Inglis, Kalshi’s head of growth, told Ad Age earlier this month that almost every brand had used AI. He said that was why the company had moved away from leading on the technology and toward a maximally human approach.

I think the Choi ad makes that positioning look less like a strategy than a communications problem. Kalshi can commission expensive human-made campaigns and still allow an agency to treat a real person’s face, race and video as editable advertising ingredients. The issue is not only that the ad was generated with AI; it is that the company’s stated human approach apparently did not extend to the person whose work it used.

The most important detail may be what the announcement leaves out: whether Kalshi knows how widely this practice was used. Saying that one old ad has been discontinued addresses the visible example, but not the system that produced it. If other creators were affected, the company’s response will show whether this was an isolated failure or a larger gap between its marketing message and its controls.