Using AI and robotics to stamp out art fraud
If you follow what we do here at Acrylic Robotics, you know we’re building tech that artists can harness to reproduce their works at scale using real paint on canvas. But you might not know that our robots are now also being tapped to help identify possible art forgeries and prevent fraud.
In early July 2025, Norval Morrisseau—widely considered the greatest artist Canada has ever produced—made headlines (again) when a third dealer charged in the Morrisseau art fraud case pleaded guilty.
The Morrisseau fraud case is complex, wide-reaching, and dates back decades. Multiple organized forgery rings produced thousands of fake replicas of Morrisseau paintings, ultimately inflicting losses of up to CAD$100 million on the estate. As many as eight fraudsters were arrested in 2023, and some of their cases are still pending. (For details, we recommend reading “The ‘Multi-Multi-Multi-Million-Dollar’ Art Fraud That Shook the World,” published in The Walrus last year.)
As far as the estate is concerned, the fight isn’t over—and this week, it announced plans to bring AI into the ring.
The estate is partnering with Acrylic to train an AI model to help detect Morrisseau forgeries and restore his legacy and the value of the estate.
After Morrisseau died in 2007, his estate began assembling an extensive catalogue of his works upon which to train an AI model. Acrylic Robotics is supporting this work by having its robots reproduce Morrisseau pieces for the model—called NORVAL AI—to use as training data.
By analyzing unique characteristics like stroke length, pressure, speed and more, our robots can create high-fidelity reproductions of artworks. To detect the possibility of forgery, NORVAL AI compares our reproductions of Morrisseau’s works to authentic pieces verified by the estate. The AI can identify discrepancies in brushstroke patterns and techniques and find deviations that indicate possible forgery.
A “game-changer” for assessing authenticity
It may seem ironic that a key plank in the plan to address forgery is the deliberate production of, well, forgeries. But these robot-made pieces have been requisitioned by the estate—and if all goes well, the project could provide an authoritative way to distinguish impostor pieces from those that have a high probability of being real.
The creation of the pieces will also help train NORVAL AI, in turn helping Acrylic create better replicas, contributing to a self-feeding improvement cycle.
To understand the significance of this emerging technology, it’s important to understand how paintings have historically been assessed for authenticity.
Provenance—the documented history of an artwork’s ownership, often in the form of gallery records, catalogues, auction sale receipts and other paperwork—was once considered the gold standard. But the courts have come to see provenance as increasingly vulnerable to manipulation because it can be falsified. Courts now expect multiple layers of verification in addition to provenance, such as stylistic and material analysis, forensic testing and third-party expert opinion.
Experts in Morrisseau’s works, like the Norval Morrisseau Estate’s executive director, Cory Dingle, can often distinguish fake paintings from authentic ones through visual inspection alone. But in Canadian law, an expert’s word is not enough—and on top of that, you can be sued for defamation for publicly raising concerns about a work’s authenticity.
Long story short, the traditional art world is not going to accept the word of one expert on which paintings are authentic, says Acrylic’s co-founder and CEO, Chloë Ryan. But an objective, third-party AI model that has proven again and again that it can establish with high probability whether art is fake or real could be a valuable tool to back up an expert assessment and establish authenticity. Although ultimately only a court ruling can serve as the final word, NORVAL AI has demonstrated a 100 percent success rate in testing so far.
“We think AI could be a game-changer for art authentication," says Ryan.
Although other tech-based authentication methods exist—such as X-ray fluorescence (XRF) scanning and infrared reflectography, these methods are costly, time-consuming and not available to the average individual. NORVAL AI can be a cost-effective first line of defence by identifying works of art that have a higher probability of being fake and could benefit from additional analysis.
“We must get beyond connoisseurship and into empirical evidence,” says Dingle. “We have to help provenance with probability. We need to expand the tools in the tool box, not only for law enforcement, but for our wonderful institutions, galleries and the general buying public.”
Designing art to be fraud-proof
A logical next step is to ensure there will be no more forgeries in the future, and Acrylic Robotics is already planning a solution for that. Our idea draws on infrared reflectography scanning, which can detect carbon materials—found in pencil, ink or graphite, but not paint—hiding beneath paint layers.
Police forces and galleries already use XRF and infrared reflectography scanners to assess paintings. Acrylic’s team will be drawing on this concept to ensure that our works have hidden indicators that identify the work as an authorized Acrylic Robotics piece, produced as part of a limited edition set.
“As one in a limited series of robot-produced pieces, it would have value—but it would be impossible to pass it off as the original,” says Ryan. “It’s extremely important to us that our reproductions don’t contribute to forgery problems.”
These and other authentication approaches are still in development, and we cannot disclose the details yet. Anyone with questions is welcome to contact the estate.
For Ryan, what’s most exciting is the enormous potential she sees in the technology.
“Hopefully we can be useful in helping to train NORVAL AI to detect the probability of fraud,” she says. “But this is really just the tip of the iceberg for the art world.”
