Isaiah 1:17 says, “Learn to do right; seek justice.” That has been Compassion’s approach to child protection and safeguarding from the beginning. We are committed to continually learning, speaking up for the vulnerable and pursuing justice whenever wrongdoing is identified or even suspected. That commitment drives our ongoing investment in people, processes and technology designed to strengthen protections for every child we serve.
Protecting children is, and will always be, our highest priority. Compassion has and continues to develop an ever-evolving comprehensive safeguarding ecosystem that includes child protection policies, mandatory trainings, screening processes, correspondence reviews, reporting mechanisms, investigations, ongoing monitoring and dedicated child protection professionals throughout our global ministry.
Recent media reports paint an incomplete and misleading portrayal of Compassion’s long-standing child protection and safeguarding efforts. In a commitment to transparency before and after publication, Compassion International provided the journalist with detailed information about Compassion’s entire safeguarding and child protection ecosystem, including how every sponsor letter is reviewed before delivery. Sadly, much of that information was not included in the articles and most of the responses were buried in a link later in the first piece.
Safeguarding is like a tapestry that grows stronger with many layers woven together. One of those layers is a Compassion-developed AI model, called PATCh (Protect All The Children)—an innovative, proprietary technology used by Compassion International to screen people and content connected with participants registered in our programs.
PATCh is not used by Compassion Canada. The model has components still under development and is designed to complement Compassion U.S.’s existing systems by providing an additional layer of protection that identifies potential risks for human review.
One of the functions of PATCh is to analyze letter correspondence and flag potential content concerns for further review. These alerts are not determinations of misconduct—rather, these alerts help identify communications that may warrant additional review by our safeguarding experts. Human review is still ongoing. We have not abandoned the human review process for AI and do not intend to do so.
The media reports blur the distinction between potential alerts flagged by an AI tool in development and confirmed safeguarding concerns that require follow up action. It’s easy to look at the numbers being used and think Compassion sponsor correspondence is harming children by the thousands. But that’s not true. Flags are not harms. They are an AI model asking a human to look at something to determine if there’s anything harmful there.
The reports would have you believe that if the AI model flags it, no further review is necessary; it must be true. But we know that’s not true. AI is not perfect. In this case, the AI model is still in development. Letter content that triggered an initial alert during Compassion International‘s recent AI test may be, and in most cases are, phrases that are appropriate and supportive in the context of a sponsor/participant relationship (e.g., sponsors telling their sponsored child “Jesus loves you and so do I.”).
Right now, Compassion International teams are reviewing a sub-set of the flagged examples that will help refine the model to more accurately flag the types of content that warrant further attention. A flag is the beginning of a review process. Not the end of one. Acting on the flags alone would be irresponsible and could harm both children and sponsors.
Furthermore, Compassion International conducted the AI test using a set of old letters sent between 2017 and 2025. The test did not include letters from Compassion Canada. When those letters were originally received, they went through the child and youth safeguarding processes in place at the time. Any letters identified as potentially harmful during that period were addressed through those established processes.
Compassion Canada’s framework
In Canada, safeguarding is supported by additional oversight, accountability and communication review processes designed to protect children and participants at every stage of the supporter relationship. Whenever concerns arise, they are escalated through a clear reporting and decision-making process that prioritizes child safety above all else.
Key elements of Compassion Canada’s safeguarding framework include:
- Participant safeguarding committee: A dedicated cross-functional committee provides shared accountability and oversight for reviewing inappropriate supporter communications or interactions. The committee has authority to recommend actions up to and including sponsorship cancellation, with child safety serving as the overriding consideration in all decisions.
- Volunteer letter review and escalation process: Trained volunteers review physical supporter correspondence using established safeguarding guidelines. Content that raises concerns is escalated to trained staff and, when necessary, to the Participant Safeguarding Committee for further review. Communications that do not meet safeguarding standards are not delivered to children.
- High standard of conduct: In addition to prohibiting inappropriate behaviour, Compassion Canada requires sponsors and visitors to avoid even the appearance of inappropriate conduct. Sponsors are asked to consider how their actions may be perceived and to maintain appropriate boundaries in all interactions with children and youth. (This requirement exceeds what is commonly found in many child protection policies and reflects our commitment to prevention, not just response.)
Another report by the same media outlet alleges that Compassion International ignored warnings regarding the potential harmful use of child photos on our online presence. On the contrary, internal discussions regarding AI-enabled image manipulation began when these risks first emerged and continue today.
The article itself recounts various conversations among staff members going back three years, discussing mitigations as early as 2024, and ongoing discussions about technical protections throughout 2025 and 2026. That’s not an organization ignoring a warning. It’s an organization prioritizing children’s safety while navigating a rapidly evolving AI landscape. Based on Compassion International’s review, we have no reason to believe any images from the website have been manipulated in this way. The article describes a potential risk, not a known misuse of our images.
Anyone who publishes any photo online is vulnerable to bad actors using those images without permission. Nonprofits, businesses, schools, churches, charities and families around the world are navigating the same challenge. There is no foolproof solution to date. Since 2021, Compassion does not share children’s full names online and we continue to test and investigate several options for deterrence and mitigation.
One of the most important safeguards we have implemented to reduce a child’s exposure to the risk of AI-enabled image manipulation is separating a child’s image from identifying information. We have and will continue to adjust how we share children’s images as the risks—and the tools to address them—develop.
We recognize that child protection work requires constant vigilance, humility and continuous improvement. That is why Compassion continues to invest in dedicated people, robust processes, specialized expertise and innovative technology designed to strengthen safeguards and respond to emerging risks. In fact, the real story here is of a 75-year-old organization that is thoughtfully exercising due diligence as it invests in, innovates and pioneers in the child protection and safeguarding space.
We welcome thoughtful conversations about child protection and believe such conversations are best served by complete and accurate information. Thoughtful conversations are not one-sided. Compassion remains fully committed to identifying, learning, responding and adapting—doing everything reasonably possible to safeguard every child entrusted to our care.
Micah 6:8 tells us to “Act justly…” (we take child protection seriously and investigate every concern) “…and to love mercy…” (we remember there are children, sponsors, staff and communities on the other side of these conversations) “…and to walk humbly with your God” (we acknowledge that safeguarding requires continual learning and improvement. We carefully and prayerfully seek God’s guidance as we listen for His voice, follow His leading and faithfully steward the children He has entrusted to our care).
