The AI backlash targeted three stills, not the whole film

The shot

'Late Night With the Devil,' a horror feature written and directed by Australian siblings Cameron and Colin Cairnes, premiered at SXSW in March 2023 and was released theatrically by IFC Films and Shudder on 22 March 2024. The film, shot as if it were surviving broadcast tape of a fictional 1977 talk show, credits David Dastmalchian in the lead role alongside a cast and crew detailed in Variety's SXSW review. In the days before wide release, viewers on Letterboxd and X began flagging what they identified as AI-generated imagery inside the film, and the directors issued a statement to Variety on 21 March 2024 to address it directly.

What the documents show

Variety's report quotes the directors' full statement: 'we experimented with AI for three still images which we edited further and ultimately appear as very brief interstitials in the film,' credited alongside 'our amazing graphics and production design team.' This is the filmmakers' own account, naming a narrow scope: three images, edited by hand after generation, used only as brief transitions. The same article traces the backlash to a 19 March Letterboxd review that said the AI content undermined the reviewer's enjoyment of 'the amazing performances and clever ending,' which then spread to X. Variety's earlier SXSW review, filed a year before the controversy, records no mention of AI at all, only credited cinematography, editing, and design work by Matthew Temple and the Cairnes brothers themselves.

The workflow

As the directors describe it, the AI-generated stills were one input among several handled by a credited graphics and production design team building a 1970s aesthetic: the images were generated, then 'edited further' before being placed as brief transitions between scenes, not used as principal photography or as a stand-in for any performance. The rest of the production followed a conventional path the SXSW review documents in full: a written screenplay, a cast of credited actors, a director of photography, and an edit assembled ahead of a festival premiere more than a year before the images in question drew attention.

What the tool does not change

Nothing in either document suggests AI touched the film's performances, its editing, or its story; the directors' statement is specific to 'three still images' functioning as interstitials, and the SXSW review's full crew list shows conventional design and camera work carrying the bulk of the film's period look. The controversy is a case study in scope: a narrow, disclosed use of a generative tool became, in social-media discussion, a claim about the whole film's visual effects, a claim the directors' own account does not support.

• Does a public statement specify exactly what was generated, or only that 'AI was used' somewhere in a production?
• Did the criticism precede or follow the filmmakers' own account of the scope of use?
• Would the same reaction have occurred if the same images had been drawn by hand instead of generated?

The two Variety pieces, filed a year apart, together separate a credited, festival-reviewed production from a narrowly scoped and later-disclosed use of generative stills, a distinction the initial social-media reaction did not make.