|
We attended the INHOPE Summit—or attempted to—to gain clarification on some of these issues. How many hotlines process images without human verification? How many hotlines accept reports of cartoon images depicting fictional minors, that don't involve the abuse or exploitation of any real child? How can stakeholders participate in the development of "baseline" factors used to classify images that are deemed illegal worldwide? How uniform are the national baseline standards applied by INHOPE members, and where are these standards published?
In the end we were unable to address any of these questions to the meeting, because a NCMEC representative complained about our Executive Director's live tweeting of the event, even alleging that one tweet was defamatory. Unfortunately this isn't an isolated incident; our previous repeated attempts to communicate with NCMEC about their policies and practices, dating back to September 2018 (the month after our launch), had also gone unanswered.
If a meeting is intended to be held under conditions of confidentiality, it's simply good practice to make that clear at the outset. Another option—the one that we adopted for our #SexContentDialogue last month—is to allow the meeting participants to quote from discussions at the meeting, but not to identify the speaker; this is called the Chatham House Rule. In Violet Blue's recent report for Engadget about our meeting, she expressed surprise that we had allowed press to attend our meeting at all, given that "usually such meetings… are kept behind closed doors."
No such stipulations were made about the INHOPE Summit, and given the public importance of the issues being discussed, we took it as our duty to report key points made at this meeting. So it came as a surprise when our representative was asked to leave the meeting over a tweet summarizing the presentation of a case in which a teenager was imprisoned for 31 years without parole for abusing his siblings. The tweet was deleted on request as a courtesy, although legal advice that we have since obtained indicates that there was no substance to INHOPE's allegation that it was defamatory of the speaker.
Other organizations in this sector have treated us in a similar manner. We have had no response to our repeated outreach to Thorn, an organization that Violet Blue recently investigated for its involvement in the surveillance of sex workers. Project Arachnid refused our request to evaluate the software that they use to automate the removal of material detected as being illegal.
As a Pro-Publica investigation revealed earlier this year, child pornography prosecutions have been dismissed due to the refusal of the secretive Child Rescue Coalition to give up details of its Child Protection System software, and our requests for access haven't been any more successful.
Google and Microsoft also maintain software used for filtering of illegal images (Google's includes a system that relies on artificial intelligence to identify new images of abuse). Although these systems are made available to other companies, our requests for access to evaluate these systems for their effectiveness and compliance with human rights norms have been refused, when they haven't been simply ignored.
To be clear, we are not singling out any one company or organization as being particularly intransigent here; the lack of transparency in this sector is a long-standing, systemic problem. The Internet Watch Foundation, which has been exceptionally open and helpful in their dealings with us, at least explained their reasoning for refusing us access to their URL blocklist and image hashes, "as both are heavily restricted due to the nature of the content contained within them."
But this reasoning doesn't stand up. Real images of child sexual abuse cannot be derived from image hashes. Neither are URL lists illegal in themselves; it is possible to scan such lists for mismatches without bringing up any illegal content. And when such lists have been leaked or reverse-engineered in the past, they have been found to contain completely innocent content.
And therein lies the problem: without independent scrutiny of these lists (even under a non-disclosure agreement, which we have indicated we are happy to sign), watchdog groups like ours have no way to verify that the sector is doing what it claims to be doing. We are not interested in naming and shaming actors, but we are calling on the sector as a whole to do better.
Our requests aren't arbitrary. We have been up-front about our intentions: we are asking for information and access to these tools because we are compiling an inaugural whitepaper on the transparency and accountability practices of Internet platforms and agencies when it comes to child protection.
Additionally, we have a practical need for access to these resources; we moderate a web forum of our own, and we partner with an independently operated support forum called MAP Support Chat. The latter had been previously censored by Internet company Discord, despite professionals attesting to the importance of such forums in helping to prevent abuse. Both of these are real-world environments for putting these systems and blocklists through their paces.
Our ejection from the INHOPE meeting today is the lowest point of what has been a poor showing of support from the sector for our evidence-based, human rights focused, and sex-positive approach. It also marks a point of divergence between the hotlines and the actual experts who are the strongest supporters of this innovative and inclusive approach.
At this point, the gloves are off. With or without cooperation from the large, government-funded Internet hotlines, we will continue to do what we believe is right, by shining a light into this largely unaccountable sector, and holding them to a higher standard of practice befitting the importance of the work that the public entrusts to them. If you believe in our work, you can click the button below to donate to our campaign to raise funds for the completion of our transparency whitepaper. |
