By Elizabeth Broadbent

Congratulations to the overseers of the HWA newsletter: this August’s issue likely had the highest open rate ever. Accolades end there, since members only opened it after they heard that it included a pro-AI screed penned by trustee Sèphera Girón (she sells her own AI-driven “writing app,” Writer’s Discipline Coach). Her piece decries “gatekeepers,” or “self-proclaimed people who have decided that the rest of the world must follow their own moral code,”  as well as what  she calls “witch hunts.” Girón never quite explains those, leaving the reader to assume they simply involve torch-bearing horror writers accusing random writers of using generative AI. 

But the targets of these witch hunts are far from random. Neither is the use of the term “witch hunt.” By design, generative AI is a technology that benefits those in power. We live in an entrenched system of capitalism; capitalism is a system of patriarchy. The backbone of patriarchy is white supremacy. Generative AI benefits a very small cabal of rich white men at the expense of marginalized people, and the marginalized people who lose the most are Black women. 

 In this particular Crucible metaphor, Girón isn’t Tituba. She’s barely even Abigail.

Generative AI: Who Loses?

Capitalism likes to think of a world in which some people win. Some people lose. The world is a binary: men/women, good/bad, white/black. This technology is not neutral. Instead, it enhances bias; we see this when Elon’s Grok starts hallucinating racist insanity. According to a Congressional white paper, Exploring the Impact of AI on Black Americans, generative AI models “amplify and exacerbate existing societal biases.” Because racism exists, generative AI makes it worse by producing text that caters to that bias. The AI picks up on existing bias and then plays to that bias because it sees that bias already prioritized. 

AI performs the same operation with positive attributes. People don’t like texts that murder puppies; therefore, AI doesn’t often generate text that murders puppies. But when AI sees that subtle bias against Black people underpins social interaction, it replicates and exacerbates that. The largest study of AI algorithms to date found evidence of clear racial bias against Black applicants. 

LLMs and LIMs (large image models) were trained on the work of uncompensated creators. Rich white men have a long, long history of stealing labor without compensation. As a group, white Americans have privilege and have had the societal opportunity to build generational wealth, both of which offer a buffer against the theft of their creative work. Black and Brown Americans have had neither. 

Moreover, the theft of their work only feeds the beast. The unique characteristics of marginalized peoples’ art—rich with cultural markers, allusions, and heritage—are elided as LLMs drag their training material to a whitebread homogenization. Everything that makes that art different and exciting disappears in the belly of the beast. Not only has that labor been stolen to produce profits for rich white men, but the nature of LLMs also strips and homogenizes it. Therefore, their uncompensated labor is used to bolster a tool that only contributes to their continued oppression. 

While all marginalized people lose out, Black women stand to lose the most. Black women have already lost the most jobs as a result of AI disruption in the workplace; AI-based facial recognition systems have an error rate of around 1% for white men, and 35% for dark-skinned women. Resumes with “Black-sounding names” or “affiliations with Black organizations” are penalized by AI hiring models. Asked to depict a “black woman smiling,” one artist found that some LIMs (large image generators) throw up not realistic faces, but surrealist nightmares (asked to depict Dakar, the same LIM threw up pictures of “arid desert landscapes and ruined buildings”). Black women have had the least opportunity to gather the generational wealth that might serve as a buffer against stolen work, and it’s no shock that any optimism Black women have about AI, according to the Brookings Institute, “is tempered by a deep-seated mistrust of the institutions deploying it and is a direct consequence of lived experience.”

Witch Hunts

We’ve seen the direct consequences of that mistrust in bookish culture. The so-called AI “witch hunts” that Girón decries are certainly happening. Witches come from the margins—the outcast, the unspeakable, the scapegoats. When we speak of witches, we speak of Tituba, the enslaved woman Abigail Williams accused in The Crucible. Witches are those society keeps on the edges,  those society prefers to remain silent. 

Reality bears this out. Most recently, Nigerian-born author Jerry Falade lost a $2 million publishing deal after his representatives said they “couldn’t verify whether AI helped him write the manuscript,” which he vehemently denies. Mia Ballard’s Shy Girl, a self-published success, was axed by Hachette after booktokers made allegations of AI use, which she also denies; H.M. Wolfe was caught up in controversy after an AI-detection tool Panagram (AI itself) reported the text as 60% AI-written. 

Training LLMs on our texts, then asking the same machines to tell us if our books were generated by artificial intelligence seems like rank tautology. We’re coming for people based on nothing but vibes, and the vibe in America has always been racism. It’s not a shock that these witch hunts’ main targets have been marginalized writers. Capitalism figures marginalized people as consumable, and always has. 

Neither is it a shock that this happens at a time when booktokker discourse has begun saying, “I don’t read books I can’t relate to” as code for “I don’t read books with Black characters.” It’s nothing more than a stunning denial of Black humanity, fascism as book discourse. No wonder Black writers are being review-bombed with single stars on Goodreads and Amazon. Generative AI is flattening reading culture so much that some whites believe they have a right to erase Black people from their narratives. 

See prime example, Elon Musk, one of AI’s biggest proponents and profiteers. He’s promised a whites-only, full-length version of The Odyssey by the end of the year. 

The HWA Speaks Out

The HWA acted decisively and appropriately by apologizing for Girón’s piece and its numerous issues. Both queer and multiracial, Mae Murray, author of “I’m Sorry If I Scared You,” points out on Instagram that when Girón falsely equates the lack of clean water on Indigenous reservations to the use of AI, she not only uses indigenous people as props “to justify her pro-AI stance,” but she also “minimizes the very real work activists are doing to fight back.” This false equivalency is doubly offensive, and the HWA acknowledged it. They affirmed the Horror Writers Organization’s stance against generative AI and removed Girón as a trustee. 

It’s a start. But we need to see better protections for cover artists; books with AI-generated covers remain eligible for Bram Stoker Awards. And while the community’s outcry against Girón was heartening, the link between generative AI and racism needs further acknowledgment and examination. Plagiarism machines reproduce prejudice, and the backbone of capitalism is white supremacy. 

We can’t stop the witch hunts until we acknowledge it.