Art Museums' AI Push Primarily Serves Tech Industry, Report Finds

(MENAFN- USA Art News) While many art museums and image libraries have embraced generative AI and partnerships with tech companies, a closer look reveals that these collaborations primarily benefit the technology sector, not necessarily the public or the art world as claimed. These initiatives are often promoted as ways to make art more“accessible” and to“democratize” it, arguments that warrant scrutiny.

Recent examples highlight this trend: the Harvard Art Museums' marketing team experimented with AI to engage new audiences, and Boston Globe reported on local museums using AI to“bring history to life.” The Prado also issued a press release, widely covered by various media, stating that an AI -powered video re-imagining its collection under solar eclipse lighting was an outreach effort for non-traditional museum visitors. However, critics argue these claims of accessibility and democratization obscure the true motivations behind such partnerships.

Contrary to the idea that museums are proactively seeking AI solutions to address declining attendance and budget shortfalls, a longer historical view suggests the tech industry has been cultivating AI usage in cultural institutions for over 15 years. AI developers initially sought vast datasets to train image recognition models, with museum collections providing a rich trove. The complexity of paintings, with their subtle color gradations, presented a compelling challenge for computer vision specialists, promising to enhance models trained on these artifacts. Academic computer scientists have explored using art for computer vision training since the early 2000s, with open-access art reproductions available as datasets on platforms like Github.

Google's Arts & Culture Platform was partly launched with such training in mind, offering access to labeled image datasets from museum collections for development. In 2022, Google reorganized Google Labs to encourage the discovery of new applications for algorithmic models trained on its Arts & Culture data. However, the data from historical images has also been applied in less benign ways. A 2025 Nature paper by six authors, titled“Computer-vision research powers surveillance technology,” highlights how data, including digital art reproductions, is used to develop models for visual policing, such as facial and body-part recognition.

For instance, in 2025, the Prado partnered with AI firm Sherpa to count crowds in paintings. While the Prado claimed this would help people“better understand” the artworks, the specific mechanism for this benefit was not detailed. This collaboration likely serves to refine computer vision for quantifying crowds, such as protest sizes, or to normalize this technology as a seemingly harmless activity. AI's determinations of criminality, personality, and pathology are opaque to the public and beyond audit, posing a fundamental democratic concern against this industry's expansion into museums. Research in computer vision is also being applied to medical diagnoses based on images, and to vetting job candidates or assessing emotional states through“affect recognition.”

The financial incentives for the data industry are substantial. Tech companies like Waymo and Uber generate revenue not just from their services, but by using their platforms and vehicles as data collection sites, gathering demographic information, audio recordings, and user reviews. From this perspective, museums are akin to Waymo, functioning as contained spaces where large numbers of people enter with diverse needs and desires, generating immense data. In 2021, Italian museums partnered with tech firms to measure visitor time spent on paintings and their expressions. Many museums are now commissioning chatbot docents for interactive tours based on their collections. These tools, while seemingly developed with museum input, ultimately serve the mercenary goals of the tech firms, either by promoting AI as a superior educational provider or maximizing data capture.

The claim that AI democratizes art, implying it makes art accessible to more people, also faces challenges. A Merriam-Webster definition of accessible refers to something“capable of being understood or appreciated.” True accessibility would involve investing in art historical education, hiring more museum educators, and avoiding layoffs, as seen with the Museum of Fine Arts in Boston. It would also mean offering opportunities to new artists and lending works to smaller or remote museums, as well as making admission free, a step the Harvard Art Museums have already taken. Instead, AI is often used to create barriers and misrepresent historical artifacts. Examples include creating interactive illusions with 18th-century figures, a chatbot Frederick Douglass narrating his life at the Museum of African American History in Boston, and the Prado paying a studio to create a video depicting paintings in lighting conditions they were not intended to portray. These“gimmicks,” while seemingly harmless, exploit culture to advertise AI, conditioning the public to trust private tech companies with cultural education. Ultimately, these collaborations grant greater access to the AI industry, not to the public, undermining true democratization.

We previously reported:

A new survey by the American Alliance of Museums reveals that 91 percent of US museum visitors have concerns about AI use in exhibitions.

Source: ARTnews

MENAFN05112026005694012507ID1111922357