Every day, people wake up trapped in forced labor. The annual profits extracted from their exploitation amount to $236 billion for illegal actors, like human traffickers and illicit organizations. Yet, for most companies sourcing goods from global supply chains, this crisis remains almost entirely invisible.

Forced labor hides in the deep tiers of supply chains within the factories, farms, and mines that many recognizable brands never visit and their auditors rarely reach. As consumer expectations and regulatory pressure mount, however, the gap between what companies claim to know about their supply chains and what is happening on the ground has never been more consequential.

A joint research initiative between the Dynamic Sustainability Lab (DSL) at Syracuse University and Thomson Reuters has spent the past three years investigating this gap. Their April symposium in Washington, D.C., and the corresponding research report, brought together customs officials, labor advocates, sustainability leaders, and technologists to confront that the tools which many organizations have relied on, but are not working — as well as the tools that could work, but are not yet being used.

The scale of forced labor is expanding. Since 2022, the list of goods identified as carrying forced labor risk has grown to 204 products across 82 countries, from 159 products across 78 countries. Indeed, the current top five types of products in which forced labor is found include electronics, garments, palm oil, solar panels, and textiles.

Equally important is dispelling the notion that the challenge of forced labor is a distant, developing-world problem. In 2024, the National Human Trafficking Hotline identified nearly 12,000 cases of human trafficking in the United States alone. Strikingly, 71% of those experiencing forced labor in the US entered the country on lawful H-2A and H-2B work visas. And U.S. Customs and Border Protection has stopped the passage of nearly $4 billion worth of goods under the Uyghur Forced Labor Prevention Act, a sign of both growing enforcement and the depth of the challenge.

Why current tools are falling short

For decades, industries have relied on social audits, codes of conduct, and traceability programs to identify and address forced labor. Each has value, but none is sufficient.

Social audits are the most widely used tool, and the most widely criticized. In 2019, ELEVATE, one of the largest auditing firms in the world, acknowledged that its own methodology is "not designed to capture sensitive labor and human rights violations such as forced labor." A review of more than 21,000 audit reports found consistently low findings on forced labor because the audits lacked the depth to find them. Companies are typically notified before an auditor arrives, and this gives them time to present a sanitized version of their operations.

Codes of conduct set labor standards in supplier contracts, but they only cover what companies can see, which is mostly Tier 1 suppliers. Research from the DSL/Thomson Reuters partnership found that 72% of Gen-Z consumers, a growing share of the market, have little or no trust that corporations are following through on stated commitments.

Traceability programs could hold real promise, but a critical statistic illustrates their current limits: Manual supplier surveys reach fewer than 10% of suppliers beyond Tier 1, leaving lower-tier suppliers largely unexamined. In the automotive sector, for example, more than 98% of aluminum tariff exposure originates at Tier 2 or deeper.

What AI can do differently

AI-powered supply chain tools offer capabilities that change the detection equation. Real-time monitoring systems use deep learning to scan news articles, social media, and government data around the clock, flagging risk signals before they surface in traditional compliance channels. Natural language processing can extract supply chain maps from publicly available online sources; and Graph Neural Networks can predict high-risk supplier relationships several tiers deep, which can connect the dots that no spreadsheet ever could.

Altana Atlas, deployed by U.S. Customs and Border Protection to support enforcement of the Uyghur Forced Labor Prevention Act, builds global supply chain maps across languages and data formats to detect evasion through complex supplier networks. In October 2025, Altana introduced Product Passports, which provide digital records that allow companies to pre-validate supply chain compliance before goods ship. The United Kingdom's Department for Business and Trade uses the same platform for its Global Supply Chain Intelligence Programme.

Supply Trace, another platform that is highlighted in the research, has cataloged nearly four million businesses, which is a volume that would have taken human researchers decades to compile.

Research data reveals a data gap

Despite this promise, adoption of these advanced tools remains low. The DSL/Thomson Reuters survey — drawn from more than 7,500 outreach contacts — found that of 73 completed responses, only 13 organizations (17.8%) are currently using AI in their supply chain operations. The remaining 82% cited a consistent set of barriers, including cost, insufficient AI-trained staff, lack of ROI or a business case, concerns about data quality, and fear that uploading proprietary information to AI platforms could expose sensitive business intelligence.

The research identifies three structural categories in which the gaps are most acute. On the data side, there are no universally accepted standards for training AI models on supply chain data, training data provenance is rarely disclosed, and suppliers deep in the chain often have no reporting data at all from which AI systems could learn.

On the workforce side, there is a critical shortage of AI-trained personnel at every level within brands, among suppliers, and inside government agencies. On the governance side, there is no central body enforcing supply chain monitoring internationally, and companies remain reluctant to share strategies around emerging technologies, even though the shared challenge is a humanitarian one.

There is also a deeper irony embedded in the research. Indeed, the workers who label the training data that powers AI systems are themselves often victims of labor exploitation. AI built to detect forced labor may, in some cases, be built upon it.

The path forward

The DSL/Thomson Reuters research points to a clear set of actions, including:

• A global data partnership against forced labor must be formed in which organizations compete not on proprietary data, but on their ability to intervene.
• Multistakeholder working groups should establish common standards for AI training data and transparency requirements.
• Human oversight must be embedded at every level of AI implementation; no algorithm should make consequential decisions about labor risk without human judgment in the loop.
• Investment in AI training  for corporate workforces, supplier networks, and government agencies alike cannot wait.

Worker testimony, the research emphasizes, must ground these systems. Data quality depends on clean corporate records and the voices of workers who know what is happening several tiers below the brand name on the label.

The DSL/Thomson Reuters partnership has committed to making this survey an annual effort, with findings expanding into peer-reviewed research and white papers. What remains is the collective decision by industry, government, and civil society to use this research provided.

You can find out more about the fight against forced labor in supply chains here‍