Infrared

Of all criminal activity, human trafficking is one of the most widespread and profitable, which is why it’s believed to be the third most prominent after drug trafficking and counterfeiting (Organization for Security and Co-operation in Europe). One of the reasons for its prevalence has to do with the fact that traffickers are at a low risk of being identified, prosecuted and sentenced. In fact, the year-over-year prosecution rate has been falling. Between 2015 and 2018, global prosecution decreased by a staggering 42% (OSCE). This decrease has meant that traffickers are, in many cases, operating with impunity, preventing the justice system from deterring those responsible. 

In Canada, only 11% of human trafficking cases resulted in a conviction (StatsCan). That’s why the justice system is often described as “failing victims”. Since the crime disproportionately affects women and marginalized communities (including Indigenous populations) it is an injustice to women’s rights and a barrier to reconciliation as “there can be no true reconciliation without justice” (Calls for Justice). 

One of the reasons that justice is difficult to achieve is due to a lack of evidence. Without high quality evidence, it is nearly impossible to prove the crime beyond a reasonable doubt, as is needed in a criminal trial. Compounding this challenge is the fact that perpetrators often create misleading evidence designed to make the trafficker appear innocent if ever they are taken to court. They do this by sending text messages that gaslight the victim, prompting a response that makes the victim seem like they are not of sound mind and/or by putting assets in their victims’ names to make it look like the profits were accrued by the victim. Furthermore, the circumstances of chaos that a human trafficking victim often endures makes it hard for them to maintain relevant documentation of their own exploitation. 

Since it is challenging to prove the crime with evidence alone, the court ends up placing a heavy burden of proof on the victims and their testimony. However, according to anecdotal evidence, the courtroom is not a hospitable place for victims; lawyers and judges are often said to discount the trustworthiness of the victim for reasons of stigma and marginalization. Furthermore, there are legitimate reasons that the clarity and consistency of a victim’s testimony may suffer, such as trauma and drug use, to which the courtroom is often not sensitive. Finally, very few if any resources are provided to the victim during the judicial process; rather, the criminal justice system is designed to serve the state. As such, not only does the justice system often fail to hold the perpetrator accountable but the process can further marginalize the victim, rendering them more vulnerable and preventing others from coming forward. 

We need a solution to the problem of insufficient evidence, overreliance on victim testimony and limited notions of justice, which are built to serve the needs of the state and do not sufficiently support the victim, creating a culture of impunity and risks of re-entry into circumstances of trafficking.

There are numerous ways in which traffickers leverage technology to recruit, advertise and exploit victims. These activities leave digital traces, which can be collected and analyzed to identify patterns that amount to indicators of suspicious, potentially criminal activity.

Infrared is an AI-enabled platform that clusters online traces of victimization to support survivors in accessing evidence, bolstering their case and helping to incriminate a trafficker. 

Infrared will offer victims and survivors the capacity to access evidence associated with their own case, along with indicators that map their digital footprint to common signifiers of trafficking and exploitation. This tool can thereby give survivors a trace of where they’ve been and other incriminating information that can be used to help bolster their memory, case and access to victim-centric services. 

This platform will be managed by human trafficking survivors who can oversee access to the service, ensuring its usability and maintaining its accuracy.

In light of our team’s values of empowering victims, the tool will be accessible only to those who are survivors of human trafficking and those supporting human trafficking survivors in accessing services. This project will require strong governance protocols to ensure the data is stored securely and only authorized persons are given access to the model’s outputs. 

This victim-centric approach doesn't suffer from the risks associated with solutions that directly support the activities of law enforcement, mainly: i) risk of false positives (harm caused to sex workers and their clients due to police interventions); and ii) risk of true positives (harm caused to trafficking victims’ in cases where law enforcement worsens the situation for the individuals involved: particularly as some individuals might not yet see themselves as trafficking victims and/or do not wish to be removed from their circumstances and/or do not trust law enforcement).

This solution will be part of a multi-sectoral approach; in collaboration with survivors, lawyers, judges and service providers. 

The risks are broken down by stage in the project lifecycle.

1. Onboarding appropriate team members

Human trafficking is an incredibly sensitive field. As such, incorporating multi-disciplinary expertise from the earliest stages in the project life cycle is vital to ensuring our work is context-aware and human rights respecting. In addition to computer scientists, our project team includes: i) survivors of trafficking; ii) legal experts; iii) a criminologist; and iv) specialists in AI ethics. We’ve commissioned a report on ethical AI practices in the domain of human trafficking from the Responsible AI Institute as well as a report on relevant legal and regulatory considerations from lawyers at Norton Rose. Finally, we’ve conducted a collaborative report between our legal and ethics experts describing implementation considerations. These outputs have been infused into the values, design and development of our tool and will continue to be referenced over the course of the tool’s lifecycle.

2. Considerations in Data Scraping

Our collection and storage of data received research approval from McGill University. Furthermore, we have commissioned an extensive consultation, performed by a law firm in Toronto, about the legal risks surrounding the tool’s deployment and use (i.e. potential scrutiny or sanctions from privacy regulators or in the context of civil liability, criminal liability and cyber intrusions).  We are aware and working within the confines of multiple legal and regulatory considerations. 

3. Considerations in Data Collection and Sharing

To ensure data security the most sensitive data has been encrypted with a unique digital fingerprint (hash).

4. Considerations in Training Algorithms

Given the impossibility of using any one indicator to predict whether online data is indicative of trafficking, our algorithms were trained with “weak labels”. Weak labels prevent the model from jumping to conclusions about whether or not a cluster of data we’ve collected contains trafficking (Nair et al., 2024, AAAI) . The model is taught only to infer trafficking in cases where multiple indicators are simultaneously met. This helps to ensure the accuracy and performance of the model, reducing the likelihood of false positives and the harm that may come as a result. 

5. Considerations in Evaluating Model Performance

Given the natural skew of our dataset, which contains an over-representation of individuals from specific demographic groups like women and members of the BIPOC community, we evaluated our model for bias. In particular, we checked whether removing demographic characteristics alter the model's predictions in order to assess whether our model was more likely to identify sex trafficking when the content features a woman from  a minority ethnic group. Our evaluation results have been mixed and point to the need for new fairness metrics, a research process we are currently undergoing.

6. Ownership Over Infrared

To account for the ongoing and evolving risks associated with the product, and to maintain accountability over the product’s outputs, we are creating a committee, which will include survivors who will be responsible for overseeing the tool’s functioning and outputs. 

There are a small number of AI solutions that have been built to perform similar functions. However, in addition to these models being fine tuned on data from the United States rather than Canada, presenting out-of-distribution problems affecting generalizability, the models also differ in terms of the end users of the tool. Specifically, many AI tools in the domain are built for law enforcement agents. 

Infrared represents an attempt at changing the paradigm, putting power back in the hands of victims and survivors to achieve justice on their own terms and in the context of their own wide-ranging needs. 

As such, the solution is built to serve members of the survivor community. Specifically, survivors who are looking for information pertaining to their particular case in order to substantiate their claims. This could be for reasons of wanting to pursue criminal or civil cases, extending immigration visas, obtaining exemptions from universities or employers for failure to meet particular requirements or in order to obtain trauma-informed support. Interestingly, in addition to enabling judicial procedures, evidence can be used by survivors to achieve financial security, immigration status or restitution. 

The tool will be managed by a diverse community of survivors who will be responsible for ensuring that the tool is adhering to governance standards (i.e. only providing evidence to survivors of human trafficking or case worker representatives) and that the tool is able to service their needs appropriately. 

In order to ensure that the appropriate community is being serviced by the tool, we will be requiring those with accounts to have their identity verified. Furthermore, data will only be shared in cases where the survivor committee has vetted the individual using the tool and the data being shared. 

Our solution has been built in the context of an incentive structure that prioritizes high impact and social benefit. Specifically, “AI for Humanity” at Mila aligns itself with the principles of Design Justice (a framework for building products by and for marginalized communities to “dismantle structural inequality and advance collective liberation”). 

Two years ago, the team signed onto Code 8.7, a UN initiative that binds projects in the anti-trafficking domain to a set of principles and best practices. As part of this program, we agreed to center the priorities of victims and survivors of human trafficking. Before engaging directly, we attended a series of Workshops, training our team on how to elevate and integrate the perspectives of victims and survivors, rather than tokenistically including them as representatives of the community. This training allowed our team to engage in productive and meaningful trust-building interactions with survivors of human trafficking, which we did bi-weekly over a six-month period. Furthermore, we made sure survivors were engaged throughout the entire product development process in order for the solution to reflect the genuine needs of the community. This type of training and sensitivity has allowed our team to deliver what we hope is a solution appropriate for a community whose data is exceptionally sensitive and for whom empowerment is vital. 

In addition to bringing human trafficking survivors onto our project team, we’ve made sure to sensitize our academically-diverse team with an awareness of the role that survivors play in this process. This team includes software engineers, machine learning developers, a UX/UI designer, criminologists, legal experts and responsible AI practitioners. As a result, each of their efforts reflect reverence to and integration of the domain and the community. The team will also be hiring a committee of representatives from the survivor population in Canada to oversee the roll-out and long-term maintenance of the solution.

Further, our team is well-positioned to deliver the solution by virtue of having integrated diverse representation from many of the relevant disciplinary domains. Specifically, we’ve worked directly with legal experts, criminologists, software engineers, UX/UI designers and machine learning expertise. The machine learning research has been overseen by a computer science professor at McGill who specializes in network connections in the context of illegal markets.

Our team has also been informed by a wide array of stakeholder groups in the Canadian landscape, having met with forensic nurses, the Canadian Center to End Human Trafficking, ACT Alberta, Public Safety, Crown Prosecutors, Royal Canadian Mounted Police and provincial police representatives from Quebec and Ontario. This positions our team well to create meaningful, effective and collaborative interventions in the domain. 

Finally, the survivor community has guided our multidisciplinary team’s research and development process in a variety of ways, including: i) promoting the need for “weak label” classification; ii) providing insight into the relevant indicators of trafficking; iii) informing the search functionality; and iv) applying the tool in a wider context of evidence collection and search for justice.   

The Infrared team has been working on this solution for over 6 years, building the technical capabilities that the model incorporates, researching the governance infrastructure required, identifying appropriate collaborators and building relationships with key stakeholders. Over the last two years, the project has benefited from meeting routinely with human trafficking survivors in order to better understand the use case and how this technical research can be applied in survivor-centric ways to the development of a solution. 

As such, the machine learning and software engineering R&D on the backend has already been completed and is working, the product idea as well as the front-end design has been validated by human trafficking survivors and other possible end users. Currently, we’re in the midst of putting each of these back- and front-end components of the tool into an integrated pipeline in order to obtain further feedback from end users. 

We are applying to the MIT Solve Competition as part of the next phase in our project’s lifecycle, moving beyond the Product Development phase (determining the desirability, viability and feasibility of the tool and building a prototype) into the MVP Phase (for which we must raise funds for product-oriented support on a longer-term basis). Since the solution currently sits within an academic research institute, taking the research into the production realm will require product-oriented funding and support, particularly from external funders. 

Funding will allow us to pilot and grow the Infrared solution, hire a committee of survivors to oversee the governance mechanisms, support our data servers/storage costs, and support with obtaining any further legal input that’s needed. 

By virtue of our positioning as one of Canada’s eminent AI research institutes, for which tech transfer is a primary goal, we have the potential to multiply project funds with federal and provincial matching programs. This funding structure will enable us to continue staffing our team with the required resources to build it once we’ve received funding through the MIT Challenge. 

In addition to funding, we are hoping to benefit from the MIT Solve network and possible in-kind support to empower the soundness of our business strategy, including our business model and financial structures. 

Unlike traditional solutions in the domain, our project puts the power directly in the hands of human trafficking victims and survivors, empowering them with their own data to use in whichever context they wish, whether to pursue criminal charges, civil cases or to be used in entirely separate domains such as with their employers, university administrator or for immigration matters. Furthermore, the tool itself has been built with, and will be managed by, members of the survivor community. This helps to ensure that trust can be maintained with our target stakeholders, enabling our work to meet the needs of end users and maintain its adaptability to the evolving landscape of online tracing. Furthermore, while current solutions struggle with accessibility and usability in the legal domain, we will ensure that the product’s outputs are explainable, interpretable and usable within a legal context.  

Another novel component of this project is that it has the potential to connect victims who have been part of the same trafficking ring to one another. Victims may be interested in connecting with others before moving ahead with a criminal case against a perpetrator on their own. This can increase feelings of empowerment, improve the likelihood of a successful conviction and aid in establishing precedent as it relates to human trafficking jurisprudence in Canada. 

From a technical perspective, the novelty of our approach is associated with our models being calibrated to the Canadian human trafficking ecosystem. In addition, our data analysis is performed in contextualized rather than de-contextualized ways. While existing literature on the topic involves decontextualized online evidence analysis, our models acknowledge the challenge in ascertaining trafficking without contextualized, clustered data. This clustered approach is helpful from a privacy and accuracy perspective. It also allows us to train the model using semi-supervised methods, an approach that requires fewer ground truth labels, which is helpful in a context where we don’t have lots of examples of confirmed human trafficking cases.

Human trafficking is a crime that is notoriously difficult to prove (UNODC). However, without sufficient proof, court cases end up relying heavily on victim testimony. This approach often leads to ”secondary victimization” and an unwillingness among victims to share information for reasons of fear, threats, or frustration with the process. All this while in a context that is not hospitable to the realities of victim’s experiences, contributing to their marginalization, insecurity and, ultimately, a failed conviction. Without justice, the courts are not playing their intended role of holding perpetrators accountable, protecting the rights and freedoms of victims or deterring others from becoming perpetrators in the future.  Not to mention, victims are not being granted their right to justice or restitution and are not being given opportunities for empowerment, which they may have sought. 

Rahel Gershuni, an international legal expert on combating human trafficking says, “in view of the typical weaknesses that plague victim testimonies, it is necessary to gather other forms of evidence and evaluate on the totality of the evidence rather than limiting it to the victim statement”. 

Infrared is built with the explicit purpose of generating new forms of evidence, taking digital traces and combining it in ways that will bolster the case against the perpetrator. This is an exciting avenue of work since, as noted by Canada’s Minister of Public Safety, “there is much potential for increased use of digital evidence in human trafficking cases despite the challenges”. Furthermore, according to Gender Rights Specialist for Amnesty International (Canada),  “digital evidence has the potential to deliver the type of overwhelming evidence that can achieve an increased number of successful prosecutions”. With greater access to evidence in a criminal case, we foresee more victims coming forward (including those who have been identified and notified as being part of the same trafficking ring by our system), more convictions, greater precedent and more of a deterrence role played by the justice system. Outside the criminal court, we foresee our tool helping victims in accessing restorative justice within family or civil court. Family and civil courts have the potential to produce verdicts that enhance economic and familial independence from the perpetrator not only for the victim but also for their children, whether in the form of financial restitution, restraining orders or sole custody. 

In the medical context, access to evidence can help to validate the need and priority of a victim when it comes to trauma-informed services, building their relationship to healthcare providers and ensuring their cases are treated appropriately. 

In the immigration setting, visa extensions, refugee cases and permanent residency applications on humanitarian grounds all become increasingly likely with substantiating evidence. 

By empowering victims with access to their own data, along with decision making power over how their data ought to be used, we’re looking to create change in how survivors rebuild their lives, putting the victims and their empowerment back into focus.

We are looking to support human trafficking victims and survivors in leveraging high quality evidence to access justice, restitution and trauma-informed services. We’re measuring progress in the context of criminal, civil and family court cases in addition to immigration, health care and student/employment outcomes. 

In criminal cases, our impact goal is the successful prosecution of traffickers and an emboldened legal landscape for the crime of human trafficking (with more clarity over the use of digital evidence). Success will be measured in terms of the cases in which this evidence is presented, the number of successful convictions and the number of future cases citing as precedent proceedings which leveraged Infrared’s evidence.  

In civil cases, our impact goal is financial restitution. Success will be measured in terms of the number of cases won by the plaintiff, the amount of money that the defendant must pay in restitution, and the number of assets recovered. 

In family court cases, our impact goal is a combination of physical freedom and financial support from the perpetrators. Success will be measured in terms of the number of restraining orders, success in custody battles and other protective measures granted to the victim. 

In the immigration context, our impact goal is to support human trafficking survivors in securing protection against deportation. Success will be measured in terms of the number of temporary residence permits and extensions issued, open work permits granted, successful refugee claims, and permanent residency applications approved on humanitarian or compassionate grounds in cases involving the use of Infrared. 

In a healthcare context, our impact goal is to enhance access to trauma-informed medical and psychological services. Success will be measured based on the number of victims who have been connected to trauma-informed healthcare services by sharing Infrared’s findings. 

From a student/employment perspective, our goal is to support victims in obtaining exemptions from Universities or employers in order to maintain their financial, immigration and social status. Success will be measured in terms of the number of exemptions granted in cases with evidence used from Infrared.

Our solution is powered by a combination of software tools and AI algorithms that work together to identify, cluster and label online data, which can be used to add credibility to claims of human trafficking. The core algorithms include: i) those for clustering (Infoshield and DeltaShield); ii) those for extracting entities, such as aliases (SWEET and NEAT) ; and iii) those for performing weak labeling prediction (T-Net). 

Clustering Algorithms

Online content that exhibit patterns can be clustered and analyzed according to indicators of human trafficking. Our experiments indicate that the algorithm, Infoshield, correctly identifies human trafficking clusters with 84% precision. We’ve also built Deltashield, which performs the same functionality as Infoshield but allows for real-time updates with incoming data, allowing our model’s outputs to not only be accurate but time sensitive as well. 

Named-Entity Extraction Algorithms

One of the valuable indicators of trafficking involves an analysis of names. However, existing Named-Entity Recognition (NER) tools deployed on online data experience challenges in their accuracy because the text can be noisy, colloquial and lacking in proper grammar and punctuation. Furthermore, the text can contain letters replaced by emojis or symbols, the writing style can be evolving and made intentionally adversarial to avoid detection. As such, most NER algorithms are not sufficient for this context. We therefore built NEAT for extracting names from online data to serve as an indicator of possible human trafficking. 

We also built SWEET, an algorithm that employs several state-of-the-art large and small language models to extract person names from the noisy ad and treat the predictions by these models as weak labels (as opposed to definitive labels) since the names can be noisy/erroneous. SWEET learns from a combination of weak labels to provide a more accurate final prediction (Liu et al, ACL, 2023).

Weak Labeling

While many tools analyze individual data for indicators of trafficking, Infrared analyzes clusters of data for indicators of trafficking. These clusters have been combined by our models because they exhibit patterns that connect them to each other. Infrared classifies these clusters into multiple classes (detecting different types of activity beyond trafficking) and uses labeling functions (LFs) to map expert labels onto the cluster with weak labels. LFs are designed by domain experts to check for indicators of specific activities (classes), namely human trafficking (HT), independent sex work (ISW) and spam (Nair & Rabany, 2023). 

We published the results of our models in top AI conferences. 

For information about Infoshield see here (published in IEEE), Deltashield see here (published in ACM), VisPaD see here (published in WebConf),  NEAT see here (published in ACL), SWEET see here (published in ACL) and T-Net see here (published in AAAI). 

There are 3 full-time staff members, machine learning Professor Reihaneh Rabbany, Senior Applied AI Projects Manager, Allison Cohen, and Software Engineer, Mark Ezema. There are 2 full-time students, PhD student Pratheeksha Nair and Masters student Vidya Sujaya; and there are 3 part-time consultants, a criminology expert, victim-engagement Advisor, and UX/UI designer. The project also benefits from legal, ethics, governance and applied machine learning team support. 

Professor Reihaneh Rabbany started the relevant technical research in 2018 after joining Mila. The applied-side research, involving legal, ethics, and domain knowledge began in 2021, with the goal of technology transfer for the academic papers.  

Throughout the project lifecycle, our team has continued to reaffirm its commitment to a specific set of values. Our values underscore the need for empowering solutions that center victims of human trafficking, a crime for which women, particularly women from vulnerable and marginalized groups, are overrepresented. In this context, we’ve sought to create a team, culture and set of protocols that favors sensitivity and inclusivity, giving all experts the room that they need to contribute and thrive. That is what has given us a strong foundation from which to ensure equity, diversity and inclusion throughout the process. 

From a diversity perspective, our team contains individuals from various nationalities, genders, lived experiences, academic backgrounds, sexual orientations, and economic circumstances. Specifically, our team consists of individuals from Iran, India, Japan, Canada and Turtle Island/Indian Country; it’s also a team being led and largely staffed by women, with male identifying individuals as our software engineer and legal expert. From an academic perspective, we have social scientists and technical experts and a range of graduate levels from  Knowledge Carriers to Postdocs, PhDs, and Masters. 

From an inclusivity standpoint, we push back against existing power structures that do not value contributions from social scientists and community members at the same level as technical experts. We do this by compensating our survivor leaders at the same rate as those with technical expertise. Furthermore, we made sure to double the hourly rate of our criminology expert once we decided she would be sitting in on the testimony of our survivor leaders. We wanted to make sure that, though she was not a full time staff, she had the flexibility to seek psychological support if she deemed it was necessary. 

In order to minimize barriers to opportunity, we interrogated the culture that we had inherited from our research institution and determined whether there were aspects that may marginalize a population or prevent them from feeling included. We decided to make our consultations with human trafficking victims highly interactive, sharing the agenda ahead of time, integrating their input and having them help steer the conversation. We also reinforced the fact that they were entitled to stop the process, not answer questions or revoke data if they would like. We praised the expertise that they brought to the project team and shared with them the role that the input had on the research agenda.

As we look ahead to the committee we plan on staffing with survivor leaders, we will make sure to have diverse representation. Since the trafficking experience can be incredibly unique from person to person, we would seek out representation among those who had been trafficked as children, those that had been nationally and internationally trafficked, those that come from different racial and ethnic groups, representatives of the Canadian indigenous communities (First Nations, Métis, Inuit) and those that are part of the 2SLGBTQQIA community (along with intersections of these identities).

Our tool provides valuable evidence that can be used by victims and survivors in the context of their criminal or civil court cases, obtaining exemptions or special accommodations from employers, university administrators, immigration authorities, or healthcare providers, and receiving trauma-informed care, among other forms of support. 

The exact value of this evidence will depend upon the institutions and circumstances in which this evidence is being used. Certain types of value can be quantified in terms of the benefit to the victim, this includes: restitution amounts, victim compensation programs ($5-25K) and salaries earned following the completion of a degree or from continued employment. There is also value derived from state-enforced security for the victim and their children and access to trauma-informed care. Other types of value, specifically those connected with the assertion of human rights and self-determination, are more difficult to quantify. However, this value can be described as feelings of empowerment, autonomy, freedom and control over one's life. These feelings may be particularly valuable among those who have been in positions of slavery, subservience and coercion. 

Value can also be quantified in terms of the value to the state. With a greater number of convictions, the justice system is serving its function of addressing serious public safety concerns and deterring future human trafficking activity. Furthermore, when human trafficking victims have access to support services that help them reintegrate, financially and economically, there is less of a burden on the state to provide shelter and care to those that are in and out of trafficking circumstances.   

By centering affected communities and connecting them with their own data, it is the victims who get to decide how their story will go on; an approach that aligns with the government’s human rights obligations, reconciliation efforts and individual aspirations for autonomy and self-actualization. 

In the long term, we are aiming to have this project funded by the Canadian government as part of their Victims Fund. The Victims Fund provides financial support to activities that (among other things) “promote access to justice and improve the capacity of service providers”. The two current priorities of the Victims Fund are exceptionally aligned with our work, specifically:

1) focus on victims of sexual violence in terms of “improving their access to justice, increase their confidence in the justice system and enhance victim services”; and 

2) victim-centered approaches to restorative justice, “giving victims of crime greater voice and choice in the criminal justice system, supporting better outcomes for victims by increasing access to restorative justice processes that are trauma-informed and victim-centered”.

This alignment is particularly promising given the results of a global study on counter human trafficking efforts which found that a quarter of existing tools are designed to identify victims and perpetrators while only 6% are built to support with victim case management and with reintegration after having been exploited. We believe this project has the potential to fill an important gap, one in which there is appetite at the federal level. 

Furthermore, our work aligns with broader ambitions within Canada’s anti-trafficking strategy, including their commitment to respond to “Calls for Justice in Reclaiming Power and Place: The Final Report of the National Inquiry Into Missing and Murdered Indigenous Women and Girls”, a report underscoring the need for self-determination, self-governance and de-colonial power structures within our justice efforts. Furthermore, it aligns with Canada’s international obligations as well. Specifically in terms of the Convention on the Elimination of All Forms of Discrimination Against Women, Convention on the Elimination of All Forms of Racial Discrimination, and UN Declaration on the Rights of Indigenous Peoples or UNDRIP. 

The evidence that this project is likely to receive support from the Victim Fund can be found in the interest we’ve already received among key stakeholders in the domain. Specifically, we’ve been invited to sit on panels hosted by Canada's Ministry of Public Safety, a large Canadian bank and a provincial law enforcement agency. We’ve also been successful in connecting with a group of Canadian lawyers and prosecutors specializing in human trafficking cases and digital evidence. This excitement from key stakeholders is likely due to the novelty of our research, centricity of our values-based approach, and reputation as being one of the world’s best in AI.