New Tools for Old Problems: Artificial Intelligence as a New Due Diligence and Authentication Tool for the Art Market?
September 20, 2023

La Belle Ferronnière (c. 1490–1496) by Leonardo da Vinci. Oil on wood, 62 × 44 cm (24 × 17 in). Louvre, Paris. This painting was at the center of one of the first litigations concerning expert opinion on authenticity.
By Dea Sula
Authentication, the act of proving that a work is what the owner purports it to be, is a cornerstone of the art market. While a buyer might appreciate the artistic qualities of a work, the connection to a particular artist is a significant generator of the value of the work, creating the difference between the original potentially worth millions of dollars and a reproduction sold for a fraction of the cost.[1] Holding out a forgery as a legitimate work can have reputational and legal consequences for buyers, sellers, and institutions in the art world.[2] The buyer’s actions during the sale may determine if they can maintain litigation due to the heightened standards art buyers are often held to as sophisticated purchasers.[3] Due to the importance of authentication in the art market, it is no surprise that authentication issues that arise during the sale process can result in potential contract failure and the generation of litigation. New artificial intelligence (AI) machines may be able to provide a new due diligence tool that protects buyers, sellers, and experts, increasing confidence in the market while giving legal protection to each player.
The Background of Authentication and Due Diligence
Authentication is not a legal concept in that the courts do not evaluate or have a standard for what qualifies as authenticated art.[4] Instead, the courts use broader contract principles such as assessing risk, due diligence, and managing liability in determining art sale authentication issues.[5] According to Black’s Law Dictionary, due diligence is
“Such a measure of prudence, activity, or assiduity, as is properly to be expected from, and ordinarily exercised by, a reasonable and prudent man under the particular circumstances; not measured by any absolute standard, but depending on the relative facts of the special case.”[6]
Further, based on Porter, a buyer has a duty to use due diligence when purchasing art, even more so if there are “red flags” that indicate something is wrong with the transaction.[7] Caveat emptor, buyer beware, is a strong concept in the American courts, especially since art purchasers are considered to be sophisticated buyers who should put care and consideration into making a large purchase such as fine art.[8] Essentially, it is the buyer’s responsibility to look into the details of the transaction in order to maintain standing in later contract litigation.
With this requirement in mind, authentication is a critical step in the sale process, as it both demonstrates due diligence and would likely reveal the “red flags” that heighten the standard of investigation a reasonable art buyer should use. Authentication is evaluated using three factors: provenance, the application of connoisseurship, and scientific evaluation.[9] Each has the potential for risk and liability, but each is necessary to establish authenticity.[10] An uninterested, third-party authentication is one of the best ways to avoid litigation issues down the line,[11] a role that AI authentication may be ideally suited to do.
Provenance
Provenance is the historical documentation that connects a work from its origin to its latest sale, establishing a chain of ownership and custody that creates a basis of authenticity.[12] A lack of strong provenance will often raise questions about theft and might even lead to the failure of a contract due to a lack of confidence.[13] While provenance is an important authentication tool, it is not conclusive proof of authenticity.[14] Issues with provenance are a significant red flag for potential buyers[15] and would likely trigger the heightened due diligence standard.
Connoisseurship
Connoisseurship is an evaluation of authenticity from an expert based on the expert’s combined knowledge of an artist’s work from evaluating hundreds, if not thousands, of examples throughout the expert’s career.[16] In a practical sense, connoisseurship is a combination of the expert’s memory from studying an artist’s work over time and their experience as the basis of their opinion. Opinions, as general ideas, are protected under the first amendment.[17] The Ollman standard highlights that there is a difference between a pure opinion, which cannot be held to be true or false, versus an actionable mixture of opinion based on fact, with art expert opinions falling into the second category.[18] Because expert opinion is a mixture of opinion and fact, experts themselves have been held liable for the opinions they give.
One of the earliest examples is Hahn v. Duveen. In this 1920s case, an art expert told a newspaper that a painting an owner claimed was La Belle Ferronnière was inauthentic, cratering its potential resale price, and a lawsuit was filed against the expert.[19] A more contemporary example is Thompson v. Andy Warhol Found. for the Visual Arts, Inc, where the board of the Andy Warhol Foundation was sued because of the authentication done by a branch of the organization.[20] The foundation spent $7 million on the litigation, followed by further litigation with the foundation’s insurance company.[21] Following this lawsuit, a variety of artist foundation boards ceased authentication altogether to avoid litigation.[22]
There is a risk of litigation when the basis of a sale is based on opinions, even more so when parties are concerned about reputational and monetary harm.[23] In art sale litigation, it is common for both sides to enlist art experts to give their opinions.[24] In these situations, experts can provide directly opposing opinions, raising significant questions about how an expert opinion may be swayed by potentially conflicting interests.[25] And if experts are held liable for their opinions, participating in litigation may create further liability, essentially creating liability for experts at every step in the process.[26]
Scientific Analysis
While less debatable than an expert opinion, scientific analysis can still generate liability and efficiency issues. The scientific analysis includes authentication done by analyzing a work’s physical properties. A popular art world concept is that testing can conclusively prove something is a forgery while proving that an artist made a specific work is almost impossible.[27] This idea is a reflection of the methods that have been traditionally used for authentication purposes. There are now a wide variety of tests to evaluate the chemical and physical elements of a work.[28] These methods are highly effective at identifying forgeries. A painting with pigments far younger than what the painter could have used is an obvious example of forgery. In a field with many opinions, scientific analysis is one method for obtaining conclusive answers.
Physical analysis requires the investment of time and money and can create insurance concerns if the work must be transported for testing. Beyond physical concerns, these methods are very adept at identifying forgeries but lack the ability to identify the work as being authentic.[29]
The Future Is Already Here
Artificial intelligence has the potential to change many aspects of society, and the art world is no different. Other industries have started using AI systems for due diligence purposes, such as bulk review of mergers and acquisitions documents.[30] Art authentication is one area where AI technology looks promising; the process may be able to produce many of the same benefits as traditional authentication methods without some of the same risks.
Artificial intelligence systems are trained on datasets and “learn” what something is based on the information the creators input into it in conjunction with data contained in the training images.[31] An example would be training an AI machine with every recognized and known work made by an artist, such as Van Gogh. Using this information, if the machine is fed another image of a work, it could identify to a specific degree of certainty if it were made by Van Gogh based on its prior training and a percentage chance of it not being painted by the artist.[32] This scenario already occurred in 2019 with the Swiss company Art Recognition’s AI identifying a Van Gogh self-portrait as being 97% likely to be an authentic Van Gogh work.[33]
AI analysis can look at a work much more closely than the human eye, down to the pixel level.[34] Another benefit is the variety of identification methods that can be used. Facial recognition has been used to identify whether a work belongs to a specific master.[35] Brushstroke analysis has also been used to identify work similar to traditional expert opinion analysis.[36] These methods have a boost of efficiency in that only photographs of the work are required for analysis rather than a physical relocation of the work, protecting the work from potential damage while saving time. AI can identify the potential for a work to be real or fake, whereas older forms of identification cannot produce a quantifiable percentage of whether a work is real.
Further, the AI machines themselves would not face personal liability for their evaluations. Whereas an expert should provide the factual basis for their opinion, an AI machine essentially provides the basis for evaluation on its face: all the data from its training set. If a machine is trained on all the attributed paintings of an artist, its percentage recommendation is derived from that analysis. An AI identification would be a mechanical process rather than an individual’s personal opinion. This process, while not infallible, removes some of the issues of bias that experts might have, including personal or financial motivations.[37] On the other hand, the owners of the AI system may guard the data set as part of protecting their investment in the machine. It is yet to be seen how these differences between a human and AI opinion will be weighted in a litigation context, especially since AI user agreements often include a clause limiting or excluding the machine’s outcome from being used in litigation.[38]
How Implementation Could Impact the Market and Legal Consequences
It would be in the best interest of all parties in the art market to utilize any tools available to increase confidence in sales and avoid costly litigation. AI system use could be beneficial before, during, and potentially after a sale occurs.
Sellers could use AI results as part of marketing a piece alongside other forms of identification such as provenance, adding assurance to potential buyers or even middlemen sellers such as auction houses. Those middlemen sellers could themselves authenticate with their own AI analysis, creating a double layer of assurance.
On the buyer’s end, AI analysis could provide another form of due diligence that gives them a more concrete picture of authenticity before finalizing a sale, potentially avoiding sale issues that appear further in the process and giving them more confidence in the security of their purchase. AI could also provide evidence of following due diligence standards by establishing that the party took the time to investigate the authenticity of a piece, meeting the heightened standards of due diligence from a sophisticated purchaser who is expected to utilize the tools at their disposal prior to concluding a sale. AI use might also constitute evidence for either party concerning red flags that would trigger heightened due diligence since a low score would indicate potential authenticity issues.
AI systems, while set up and run by individuals, do not have a profit motive in individual evaluations, a concern that might be more of an issue with individual experts that may be personally invested in a sale occurring, especially if they are being paid by one of the parties involved in the sale. AI examination may bolster expert opinions and give the experts another layer of protection from personal liability as an AI result could be used as factual evidence to support a conclusion of finding authenticity.
In a legal sense, AI use could strengthen the due diligence standard to offset potential liabilities for all parties involved in the art market. Contracts that require AI testing could become more common if AI authentication becomes a standard in the art world business practice. In effect, this might deter future litigation if all parties could do their own testing prior to the finalization of a contract. The basis of the art market is confidence, and using AI as an authentication tool could make the market more secure. As with many of the questions and potentials of artificial intelligence, only time will tell how effective this new tool will be.
Disclaimer: This and all articles are intended as general information, not legal advice, and offer no substitution for seeking representation.
About the author
Dea Sula is a current law student at the Santa Clara University School of Law with a focus on Art and Intellectual Property Law. She graduated from the University of Texas at Austin with a Bachelor of Fine Arts in Art History. She was a summer 2023 legal intern at the Center for Art Law and is the project coordinator for the CfAL Case Law Corner Project.
Further Reading
For a general overview of the AI training process: https://spectrum.ieee.org/this-ai-can-spot-an-art-forgery
Interview with the makers of the Art Recognition AI: https://www.artnome.com/news/2019/9/12/can-ai-art-authentication-put-an-end-to-art-forgery
Center for Art Law Resource on AI and Authenticity:
https://drive.google.com/file/d/1Puxw0FbU2OUu80G7wDc3vyO294v7vuAP/view
- Chiara Bastoni, Value in Art – What Makes Art Valuable?, Artland Magazine, https://magazine.artland.com/value-art/ ↑
- Caroline Mortimer, Modigliani art exhibited at Ducal Palace in Genoa revealed to be almost entirely fake, Independent (January 12, 2018), https://www.independent.co.uk/arts-entertainment/art/news/italy-modigliani-fake-show-police-investigation-art-genoa-a8154701.html ↑
- William L Charron, The Art Law Review: Assigning Burdens of Diligence in Authenticity Disputes, The Law Reviews (January 6, 2023), https://thelawreviews.co.uk/title/the-art-law-review/assigning-burdens-of-diligence-in-authenticity-disputes ↑
- Id. ↑
- Id. ↑
- DUE DILIGENCE Definition & Legal Meaning, The Law Dictionary, https://thelawdictionary.org/due-diligence/#:~:text=Such%20a%20measure%20of%20prudence,facts%20of%20the%20special%20case. ↑
- Porter v. Wertz, 68 A.D.2d 141, 416 N.Y.S.2d 254 (1979), aff’d, 53 N.Y.2d 696, 421 N.E.2d 500 (1981). ↑
- ACA Galleries, Inc. v. Kinney, 928 F. Supp. 2d 699 (S.D.N.Y. 2013), aff’d, 552 F. App’x 24 (2d Cir. 2014). ↑
- Ronald D. Spencer, Art Law on Protection for Art Experts, Artnet News (February 1, 2013), https://news.artnet.com/market/protection-from-legal-claims-for-art-experts-29980 ↑
- Leila Amineddoleh, Purchasing Art in a Market Full of Forgeries: Risks and Legal Remedies for Buyers, 22 Int. J. Cult. Prop. 419 (2015), http://authenticationinart.org/pdf/literature/Amineddoleh-Leila-Purchasing-Art.pdf ↑
- Id. ↑
- Brian Ng, Why Provenance Matters to Art Collectors, Artsy (July 26, 2022), https://www.artsy.net/article/artsy-editorial-provenance-matters-art-collectors ↑
- Eileen Kinsella, What the Story of a Botched $1 Million Auction Reveals About the Clash Between New Anti-Money Laundering Laws and Client Confidentiality, Artnet News (May 30, 2023), https://news.artnet.com/art-world/artcurial-patrick-mathiessen-aml-conflict-2309423 ↑
- Natasha Fekula, Provenance: A valuable tool to mitigate risk in fine art deals, AXA XL Insurance (February 10, 2020), https://axaxl.com/fast-fast-forward/articles/provenance_a-valuable-tool-to-mitigate-risk-in-fine-art-deals ↑
- Id. ↑
- Ronald D. Spencer, Art Law on Protection for Art Experts, Artnet News (February 1, 2013), https://news.artnet.com/market/protection-from-legal-claims-for-art-experts-29980 ↑
- First Amendment, Cornell Law School Legal Information Institute (2022), https://www.law.cornell.edu/wex/first_amendment ↑
- Ollman v. Evans, 750 F.2d 970 (D.C. Cir. 1984). ↑
- Hahn v. Duveen, 133 Misc. 871, 234 N.Y.S. 185 (Sup. Ct. 1929). ↑
- Thompson v. Andy Warhol Found. for the Visual Arts, Inc., 103 A.D. 3d 528, 529 (N.Y. App. Div. 2013). ↑
- Eileen Kinsella, A Matter of Opinion, Artnews (February 28, 2012), https://www.artnews.com/art-news/news/a-matter-of-opinion-512/ ↑
- Id. ↑
- Id. For example, a variety of experts believed 74 Degas plaster casts were inauthentic but were wary to publicly say so and met in private to discuss their opinions. ↑
- Bethany Mar, Experts’ Role in Art Authentication, Seton Hall Law (2021), https://scholarship.shu.edu/cgi/viewcontent.cgi?article=2243&context=student_scholarship ↑
- Anne Laure Bandle, Fake or Fortune? Art Authentication Rules in the Art Market and at Court, 22 Int. J. Cult. Prop. 379 (2015), https://www.cambridge.org/core/journals/international-journal-of-cultural-property/article/fake-or-fortune-art-authentication-rules-in-the-art-market-and-at-court/09D9F3F51667468E4F0B2D4BF9428652 ↑
- Amineddoleh & Associates LLC, Distrusting Disavowals: Artists Manipulating the Authentication Process (Feb 18, 2022), https://www.artandiplawfirm.com/distrusting-disavowals-artists-manipulating-the-authentication-process/ ↑
- Andrea Ouyang, The Science of Art: How scientists unmask fakes and forgeries, Yale Scientific (May 21, 2016), https://www.yalescientific.org/2016/05/the-science-of-art-how-scientists-unmask-fakes-and-forgeries/ ↑
- Id. ↑
- Jane Kallir, Art authentication is not an exact science, The Art Newspaper (November 23, 2018), https://www.theartnewspaper.com/2018/11/23/art-authentication-is-not-an-exact-science ↑
- Chris O’Leary and Raees Nakuhuda, How AI for M&A due diligence is changing every aspect of the deal process, Thomson Reuters Insights (April 16, 2023), https://legal.thomsonreuters.com/en/insights/articles/how-ai-and-document-intelligence-are-changing-the-legal-tech-game ↑
- To see the process of training and determining which data to use: Steven J. Frank, This AI Can Spot a Forgery, IEEE Spectrum (August 23, 2021), https://spectrum.ieee.org/this-ai-can-spot-an-art-forgery ↑
- Art Recognition, Vincent van Gogh: “Self-Portrait”, https://art-recognition.com/case-studies/vincent-van-gogh-self-portrait/ ↑
- Fei Lu, Introducing The Latest Tool In Art Authentication: AI, Jingculture & Crypto (September 15, 2021), https://jingculturecrypto.com/art-recognition-ai-art-authentication/ ↑
- University of Bradford, Mystery portrait is “undoubtedly” Raphael masterpiece, according to new scientific analysis, (January 16, 2023), https://www.bradford.ac.uk/news/archive/2023/mystery-portrait-is-undoubtedly-raphael-masterpiece-according-to-new-scientific-analysis.php ↑
- Ellen Wexler, Artificial Intelligence Identifies Long-Overlooked Raphael Masterpiece, Smithsonian Magazine (January 27, 2023),https://www.smithsonianmag.com/smart-news/artificial-intelligence-identifies-long-overlooked-raphael-masterpiece-180981528/ ↑
- Ahmed Elgammal, Yan Kang, and Milko Den Leeuw, Picasso, Matisse, or a Fake? Automated Analysis of Drawings at the Stroke Level for Attribution and Authentication, Proceedings of the AAAI Conference on Artificial Intelligence, 32(1) (2018), https://doi.org/10.1609/aaai.v32i1.11313 ↑
- An example of an expert using their position for personal monetary gain: Michael Day, Modigliani ‘expert’ accused of being art’s biggest fraud, Independent (January 23, 2013), https://www.independent.co.uk/arts-entertainment/art/news/modigliani-expert-accused-of-being-art-s-biggest-fraud-8463883.html ↑
- Article 6 of Art Recognition Agreement: “The Client shall not name the Examination Software or its source code and user interface as evidence in legal proceedings (e.g., proceedings before state courts and arbitration courts) and shall not call the Examiner or its bodies and employees as witnesses or experts.” https://art-recognition.com/terms-and-conditions/ ↑