The Forgery Problem

Why the art market’s greatest threat is not the forger — but the silence between the experts who could stop them

Infrastructure of Fact

By Jeff Kluge Artist, AI Ethicist, Host of Talking Art, and EVP Strategy at Authentify Art

A painting can pass through auction houses, collectors, museums, and experts — and still remain fundamentally unresolved.

Not because knowledge is absent.

But because the knowledge that exists cannot assemble itself.

There is a particular kind of confidence that has long protected the art market from its worst actors.

It is the confidence of accumulated expertise.

The conservator who has spent thirty years studying a single artist’s hand.The provenance researcher who knows every gap in the archive.The auction specialist who can read a painting’s condition from across the room.

This expertise is real.

It is hard-won.

For decades, it was enough.

It is no longer enough.

Artificial intelligence has fundamentally altered the economics and mechanics of art fraud. What once required years of skilled fabrication — a convincing fake, a plausible paper trail, a credible history — can now be assembled in hours.

Large language models can generate exhibition histories, collector correspondence, and auction records that appear indistinguishable from genuine documentation. Image-generation systems can fabricate visual provenance: photographs of works in fictional gallery contexts, digitally altered catalogue images, or simulated installation records.

The modern forgery is no longer simply painted.

It is constructed.

Synthetic. Scalable. And increasingly difficult for any single expert to detect in isolation.

But the deeper threat to the art market is not the sophistication of the forgery itself.

It is the structure of the system meant to detect it.

The Real Failure Is Not Expertise. It Is Architecture.

The art market does not lack expertise.

It lacks the infrastructure that allows expertise to accumulate.

Conservators, provenance researchers, insurers, scientific laboratories, legal specialists, and auction houses all generate valuable knowledge about works of art. Each discipline contributes insights that can confirm authenticity — or raise doubts about it.

Yet this knowledge rarely travels beyond the institution that produced it.

A conservator may identify an anomaly in a ground layer, but the provenance researcher never sees the report.

An insurer may quietly decline coverage on a painting, yet the auction house that accepts the work months later has no mechanism for seeing that signal.

A scientific analysis may raise questions about pigment composition, but the results remain locked in a private laboratory report commissioned for a single transaction.

Each expert may see a fragment of the truth.

But the fragments rarely meet.

Authentication fails not when expertise is absent, 

but when expertise cannot see itself.

The result is a market that possesses extraordinary expertise — yet operates with remarkably little institutional memory.

Every transaction begins almost from zero.

The Forgery Problem Is Now an Information Problem

For most of the twentieth century, forgery was constrained by physical limitations.

A convincing fake required period-appropriate materials, mastery of an artist’s technique, and the patience to construct a plausible provenance history. Sophisticated forgery rings existed, but scaling these operations was difficult.

Artificial intelligence has altered that equation.

Image-generation systems can produce stylistically coherent works in the manner of almost any artist with a sufficient digital footprint. Language models can generate detailed historical narratives linking artworks to exhibitions, collectors, and estates.

These capabilities dramatically reduce the cost of constructing convincing deception.

Forgery, once limited by craftsmanship, is now limited primarily by information.

And information — in the art market — remains deeply fragmented.

A Market That Knows It Has a Problem

Evidence of this shift is increasingly visible across the broader art ecosystem.

The 2025 Art & Finance Report from Deloitte found that 76% of collectors, 75% of wealth managers, and 74% of art professionals believe technology will play a central role in addressing authenticity and provenance challenges.

Yet the same report revealed that only 27% of wealth managers expressed high confidence in the quality of available art-market data.

For institutions making lending, insurance, and estate-planning decisions on the basis of that data, this gap represents more than inconvenience.

It represents structural risk.

The market understands that verification tools must evolve.

What it has not yet built is the infrastructure capable of supporting them.

A Different Approach

Authentify Art was founded on a simple premise:

Authentication improves when expertise compounds rather than remaining isolated.

The research-grade AI systems developed by Authentify Art are designed not to replace specialists, but to allow their findings to accumulate into a living evidentiary record of a work of art — one that can follow the work across transactions, institutions, and generations.

Such systems do not produce a single verdict.

They create an environment in which conservation findings, scientific analyses, provenance documentation, insurance signals, and legal assessments can be examined together.

Authentication becomes not a solitary judgment.

But an evolving body of evidence.

PART ONE

The Threat Has Evolved. The Response Has Not.

Artificial intelligence has transformed three fundamental aspects of art fraud:

1. the production of convincing forgeries

2. the fabrication of provenance documentation

3. the verification process used to evaluate authenticity

Understanding these dimensions separately reveals how dramatically the landscape has changed.

1. The Forgery Itself

Historically, convincing forgery required rare combinations of artistic skill, material knowledge, and patience.

Physical constraints limited scale.

AI has weakened those constraints.

Image generation systems can replicate stylistic patterns across large bodies of work. Robotic painting systems can reproduce brush pressure and gesture patterns derived from authenticated works.

These systems do not yet replace human craftsmanship.

But they dramatically reduce the cost of experimentation and iteration.

The result is a future in which the volume of plausible forgeries increases dramatically, placing new pressure on authentication systems that already struggle to coordinate expert knowledge.

2. The Paper Trail

Provenance has long served as the art market’s primary defense against forgery.

A work with a verifiable ownership history is harder to falsify than the object itself.

Generative AI challenges this assumption.

Language models can produce convincing correspondence, invoices, and exhibition narratives that mimic historical documentation.

These narratives often exploit the same gaps and ambiguities that characterize legitimate art-market records.

More troubling is the phenomenon of unintentional misinformation.

Collectors and researchers increasingly rely on generative systems for historical information, sometimes receiving fabricated connections to exhibitions, estates, or collectors.

Presented in good faith, these errors can enter the provenance record and propagate through subsequent transactions.

Forgery no longer requires malicious intent.

It can emerge from automated certainty.

3. The Verification Gap

The most significant vulnerability lies not in the forgery itself but in the structure of verification.

The art market’s expert community operates in a series of highly specialized but largely isolated disciplines.

Conservation laboratories produce technical reports. Provenance researchers assemble archival histories. Scientific laboratories conduct material analysis. Insurers and legal specialists assess risk and title.

Each produces valuable information.

But the information rarely accumulates into a unified record.

A system designed around sequential evaluation cannot keep pace with a world in which deception can be generated in parallel.

PART TWO

Consumer-Grade vs. Research-Grade AI

The art market’s encounter with artificial intelligence has largely been shaped by consumer-grade tools.

Image generators, chatbots, and valuation algorithms trained on public auction data have democratized access to information and introduced new collectors to data-driven research.

These tools are valuable.

But they are not designed for authentication.

Consumer-grade systems prioritize accessibility and breadth.

Research-grade systems prioritize verification and evidentiary rigor.

A research-grade authentication system must operate on different principles:

• curated and verified datasets

• probabilistic reasoning with explicit uncertainty

• auditability and traceable sources

• workflows designed for professional collaboration

These systems are not designed to replace experts.

They are designed to help experts see the full evidentiary landscape surrounding a work of art.

PART THREE

Building Institutional Memory

The most important contribution technology can make to authentication is the creation of institutional memory.

Every conservation report, pigment analysis, provenance discovery, insurance decision, or legal assessment contains information that may become relevant in future evaluations.

Today, that information rarely travels with the artwork.

It remains distributed across private archives and confidential reports.

Platforms designed as authentication infrastructure treat these findings as structured evidence rather than isolated documents.

Over time, this creates something the art market has historically lacked:

A cumulative evidentiary record.

Each assessment becomes part of a larger analytical history that improves future evaluations.

Authentication becomes more accurate not because a single expert becomes more knowledgeable, but because the system learns from the accumulated experience of many experts.

PART FOUR

Ethical Infrastructure

Authentication systems built on artificial intelligence must address three ethical dimensions.

Artists

Living artists and estates must retain agency over how authenticated works are represented within analytical systems.

Authentication infrastructure should strengthen, not weaken, the artist’s control over the historical record of their work.

Cultural Representation

Training datasets must expand beyond Western canonical markets to ensure analytical models respect diverse artistic traditions and material practices.

Authentication systems that fail to represent the full diversity of global art risk reinforcing historical bias.

Institutional Accountability

Institutions making financial or legal decisions based on authentication data must be able to audit the reasoning behind those conclusions.

This requirement aligns with broader expectations for explainability in AI systems.

Indeed, the Art & Finance Report by Deloitte emphasizes that AI systems used in institutional decision-making must demonstrate transparency, reliability, and traceable data sources before stakeholders will rely on them.

Conflicts of Interest and Structural Integrity

Authentication systems must also address a less visible but equally consequential ethical challenge: conflicts of interest in the development and deployment of analytical tools.

In recent years, a growing number of technological solutions have been introduced to the art market by organizations whose financial incentives are directly connected to the outcomes of authentication decisions. Auction houses, market intermediaries, and private stakeholders have invested in analytical tools intended to assist with verification and valuation.

These efforts have produced valuable innovations.

But they also introduce structural tensions.

When the entities responsible for determining authenticity or value also have financial exposure to the outcome of that determination, even the most sophisticated analytical systems risk being perceived as partial.

The challenge is not simply technological.

It is institutional.

Research-grade authentication infrastructure must therefore be designed with governance structures that protect analytical independence. Data sources, analytical methods, and system outputs must be auditable and transparent to the professionals who rely on them.

Equally important, the development of such systems should avoid reliance on funding structures that implicitly bias the conclusions they produce.

The credibility of authentication does not rest solely on accuracy.

It rests on trust.

And trust requires that the systems used to evaluate authenticity operate at a visible distance from the financial incentives that surround the art market itself.

Welcome to the Infrastructure Era 

Every mature professional field eventually reaches a point where expertise alone is no longer sufficient. At that moment, the discipline develops infrastructure that allows knowledge to accumulate and travel.

Medicine reached this point with the development of shared imaging databases and diagnostic systems that allow physicians to compare scans across institutions. Finance reached it with networked fraud-detection systems capable of identifying suspicious patterns across millions of transactions. Aviation reached it through global safety reporting systems that allow every incident to inform the next flight.

In each case, the transformation did not occur because experts became more capable.

It occurred because the infrastructure connecting them improved.

The art market now faces a similar moment.

For generations, authentication has depended on the insight of individual specialists — conservators, historians, scientists, and scholars whose expertise often spans decades of focused study. That expertise remains indispensable.

But the environment surrounding it has changed.

The speed with which convincing deception can now be generated — through digital fabrication, synthetic documentation, and networked distribution — places new demands on verification systems that were designed for a slower and more fragmented era.

The response cannot simply be more expertise.

It must be better infrastructure.

Authentication systems that allow evidence to accumulate across time, institutions, and disciplines represent the next phase of the art market’s development. In such systems, each conservation report, scientific analysis, provenance discovery, and legal assessment contributes to a growing evidentiary record.

Authentication becomes less dependent on isolated opinions and more grounded in structured knowledge.

This transition does not diminish the authority of experts.

It strengthens it.

By allowing their insights to travel with the work and inform future assessments, the system preserves the value of expertise while dramatically increasing its reach.

The art market is entering its infrastructure era.

And the institutions that build that infrastructure will shape the standards by which authenticity is understood for generations to come.

The Future of Authentication

For centuries, the authenticity of a work of art has depended on the judgment of specialists.

Those judgments remain essential. No system — technological or otherwise — can replace the insight developed through decades of study.

But the conditions under which those judgments are made have changed.

The speed and scale of modern information systems mean that deception can now be generated faster than traditional verification structures can respond. In such an environment, expertise alone cannot carry the burden of authenticity.

It must be supported by infrastructure.

The next phase of the art market will not be defined simply by better tools, but by better systems for connecting the people who use them. Conservation findings, scientific analyses, provenance research, insurance signals, and legal assessments must no longer remain isolated fragments of knowledge.

They must become part of a shared evidentiary record that travels with the work itself.

When that happens, authentication ceases to be a fragile process dependent on isolated opinion.

It becomes a durable system built on accumulated knowledge.

The art market has long been defined by extraordinary expertise.

Its future will be defined by the systems that allow that expertise to work together.

And in that future, authenticity will not rest on silence between experts.

It will rest on the evidence they can see together.

The future of authenticity will not belong to the loudest opinion in the room, but to the systems that allow evidence to accumulate.

Jeff Kluge is an artist, AI ethicist, and EVP of Strategy at Authentify Art.

Authentify Art builds research-grade authentication infrastructure designed to allow expert knowledge to accumulate and travel with the work.