This contribution was written by our Program Manager, Auréliane Froehlich.
Early in 2025, Wikirate partnered with Walk Free to collect data for their report examining the transparency of disclosures under Mandatory Human Rights Due Diligence Laws (mHRDD).
The result is a report authored by Walk Free, Beyond Compliance: How to Design Effective Mandatory Human Rights Due Diligence Laws, showing that while corporate disclosures still lack transparency, stricter reporting obligations and meaningful enforcement can improve disclosure quality.
It’s easy to read a report and focus on the findings, but the work behind the numbers often goes unseen. Wikirate and Walk Free analyzed 225 corporate reports – 99 under mHRDD and 126 under Modern Slavery Acts – some hundreds of pages long, turning dense PDFs into structured, comparable data. The process was complex, time-consuming, and required careful human judgment at every step.
For us at Wikirate, this project was a humbling reminder of just how challenging it is to assess corporate human rights disclosures at scale. It requires hours of dedicated work, skill, and attention to detail. And the lessons we learned along the way – not about what, but how companies report – can also help inform the design of future due diligence laws.
The hidden work: finding corporate disclosures
Before we could begin the assessment, we first had to find the reports. That alone took a surprising amount of effort. Unlike the UK or Australian Modern Slavery Registries, where companies publish their due diligence disclosures, no such government-managed repositories existed in France, Germany, or Norway. In France, civil society created a platform to keep track of reports published under the Devoir de Vigilance (the French mHRDD law), which ended up being extremely useful.
For the rest, we had to dig, company by company.
Sometimes the reports were tucked away in obscure parts of corporate websites, buried within sustainability subpages, or lost in download centers. Other times, they weren’t posted at all, and we had to puzzle out whether the company had reported somewhere else or not reported at all.
This reinforced something we say often: transparency isn’t real transparency if the public can’t easily find the information. This is why we continue to push for public, centralised, machine-readable registries.
Turning PDFs into data: why humans still matter (a lot)
After we finally found the reports, the real work began: reading them, understanding them, and turning them into structured data. We used a methodology developed with Walk Free, building on years of experience assessing Modern Slavery Statements for the Beyond Compliance Project. This part of the project is the most time-intensive. Our team, supported by 18 Wikirate contributors (many of them students or sustainability professionals), read and evaluated every report before entering the data into Wikirate.
At this point, you might wonder: why not let AI do it? We wondered too. But in practice, extracting information from PDFs is not as easy as it seems. PDF reports are notoriously difficult to parse for LLMs, especially if they contain many complex elements such as charts, tables, and infographics, and because they do not follow a predictable structure. Another challenge is that reports are full of vague, ambiguous language that requires interpretation. A few examples:
- In corporate disclosures, the line between what companies actually did and what they might do can be blurry.
- Critical information, like the absolute number of incidents or whether cases were investigated, may be missing. Documenting those absences is key.
- Terminology varies from one company to another, so pattern-matching isn’t enough.
- And crucially, human assessors spot issues with the methodology itself and help refine it.
Although Wikirate is exploring opportunities to automate parts of the extraction process, our approach remains intentionally cautious. For this project, human-centred assessment, supported by peer review, was crucial for producing high-quality, trustworthy data.
That said, we see many opportunities for companies to make disclosures more compatible with automated processing. Publishing reports in machine-readable formats (e.g., iXBRL), tagging key information, or using standardised structures (such as those in the LkSG) and strict reporting guidelines would significantly reduce ambiguity and improve comparability.
Persistent data gaps
As we analyzed disclosures, one theme emerged consistently: reports often lacked the detail needed to understand whether modern slavery was present in the companies’ supply chain and what they are doing to respond to cases of modern slavery when it was.
Most disclosures focused on policies and vague descriptions of internal processes. These are important, but they only tell part of the story. If you’re trying to understand what a company is actually doing to prevent human rights harms, or how it responds when something goes wrong, the details often just weren’t there. Some common gaps we saw:
- Incident counts with no explanation of the nature or severity of cases.
- Grievance numbers with no breakdown of complaint types or outcomes, or broken down by categories that were too broad (“Human Rights”).
- Generic references to “stakeholders” without clarifying if workers in the supply chain were ever included or spoken to.
- Generally, a lack of distinction between what was done and what is contained in policies.
When you’re trying to build a clear picture from incomplete information, you end up making judgment calls at every turn. And if it was challenging for us as researchers reviewing several statements, it’s easy to imagine how difficult it would be for workers, regulators, or civil society to make sense of these disclosures. These gaps make it harder to know what’s actually happening in supply chains.
Improving transparency through due diligence laws
One of the lessons we take away from this project is also featured in Walk Free’s takeaways: mandatory reporting alone is not enough. We need what often ends up being forgotten about: machine-readable formats, standardised structures, centralised registries, and reporting requirements that prioritise clarity and leave little space for ambiguity. Without all of this behind-the-scenes work, corporate transparency will remain limited, and the full potential of mandatory human rights due diligence laws will go unrealised.
As policymakers refine existing laws and develop new ones, and while companies grapple with expanding due diligence obligations, one thing is certain: better data is necessary if we want these laws to drive real change for the people they are meant to protect.