Content moderation is a can of worms. For Internet infrastructure intermediaries, it’s a can of worms they are particularly poorly positioned to tackle. And yet Internet infrastructure elements are increasingly being called on to moderate content — content they may have very little insight into as it passes through their systems.
The vast majority of all content moderation happens on the ‘top’ layer of the Internet — such as social media and websites, places online that are the most visible to an average user. Platforms and applications moderate the content that gets posted on their platforms every day. If a post violates a platform’s terms of service, the post is usually blocked or taken down. If a user continues to post content that violates a platform’s terms, the user’s account is often suspended. These types of content moderation practices are increasingly understood by average Internet users.
Less often discussed or understood are the types of services facilitated via actors in the Internet ecosystem that both support and exist under the upper content layers of the Internet.
Many of these companies host content, supply cloud services, register domain names, provide web security, and many more features of what could be described as the plumbing services of the Internet. But instead of water and sewage, the Internet deals in digital information. In theory, these ‘infrastructure intermediaries’ could moderate the content. But for reasons of convention, legitimacy, and practicality they don’t usually do it on purpose.
However, some notable recent exemptions may be setting a precedent.
Amazon Web Services removed Wikileaks from their system in 2010. Cloudflare kicked off the Daily Stormer. An Italian court ordered Cloudflare to remove a copyright-infringing site. Amazon suspended hosting for Parler.
What does all this mean?
Infrastructure may have the means to perform ‘content moderation’. But it is critical to consider the effects of this trend to prevent harming the Internet’s underlying architecture.
In principle, Internet service providers, registries, cloud providers and other infrastructure intermediaries should be agnostic to the content that passes over their systems. Their business models have nothing to do with whether one is sending text, audio or video. Instead, they are meant to act as neutral intermediaries, providing a reliable service.
In a sense, they operate the plumbing system that delivers the water. While we might engage a plumber to inspect and repair our pumps, do we feel comfortable relying on the plumber to check the quality of the water every minute of every day? Should the plumber be able to shut off water access indefinitely with no oversight?
Despite this, big companies have made decisions to moderate content that is clearly out of their scope as infrastructure intermediaries. It begs the question — why? Were these actions to uphold some sort of moral authority or primarily on the business grounds of public perception? How comfortable are we with these types of companies ‘regulating’ content in the absence of — or even at the direction of — governmental regulation?
If these companies add content moderation to their responsibilities, it takes away the time and resources they can dedicate to security, reliability, and new features, some of which may even help fight reasons for wanting to moderate the content. And while large companies may have the means, it adds an additional role outside of their original scope or mission that would be costly or unattainable for most startups or smaller companies.
As pressure mounts from public opinion, regulators, and courts, we should recognize what is happening and properly understand where problems can be best addressed and what problems we don’t know enough about to warrant messing with the plumbing of the Internet just yet. Moreover, we should be wary of any regulation that may turn to infrastructure intermediaries explicitly to moderate content.
Asking an infrastructure intermediary to moderate content would be like asking the waiter to cook the meal, the pilot to repair the plane, or the police officer to serve as the judge. Even if it were possible, we must ask whether it is truly an acceptable approach.
Where do we conduct moderation?
The Internet is often referred to as a layered architecture because it comprises different types of infrastructure and computer entities. Expecting them to each moderate content indiscriminately would be problematic. Who would they be accountable to?
A core idea often proposed is that content moderation should occur at the highest available layer, closest to the user. Some even argue that content moderation below this, in the realm of infrastructure, is more problematic because these companies cannot easily moderate a single content item. Infrastructure needs to work at scale, and moderating a single piece of content may mean effectively turning off a water main to fix a dripping faucet. That is, infrastructure companies often have to paint with a broader brush by removing an entire user or an entity’s access to their service.
These broad strokes of moderation are often deep and wide in their effect, and critics argue they go too far. Losing access to a system is clearly more final than having a single item removed from a system.
Georgia Evans summarized the problem well, saying ‘the deeper into the stack a company is situated, the less precise and more extreme their actions against harmful content are.’ For this reason, Corinne Cath refers to them as reluctant sheriffs and political gatekeepers. These are important complexities that must be woven into any understanding of deep-layer moderation by Internet infrastructure companies and policymakers.
The tech community and policymakers must ensure that no policy proposals unintentionally require the plumber’s legal role to include quality assurance and access determination. In the realm of the Internet, certain actors have certain functions and things work in a modular, interoperable way by design. The beauty of the Internet is that no one company or entity must ‘do it all’ to achieve a better Internet. But, we must also ensure that new demands for additional functionality — for example, moderation — are situated at the right layer and target the party with the expertise and role most likely to do a careful job.
Policymakers must consider the unintended impacts of content moderation proposals on infrastructure intermediaries. Legislating without due diligence to understand the impact on the unique role of these intermediaries could be detrimental to the success of the Internet and an increasing portion of the global economy that relies on Internet infrastructure for daily life and work.
Conducting impact assessments before regulation is one way to mitigate the risks. The Internet Society created the Internet Impact Assessment Toolkit to help policymakers and communities assess the implications of change — whether those are policy interventions or new technologies.
Policy changes that impact the different layers of the Internet are inevitable. But we must all ensure that these policies are well crafted and properly scoped to keep the Internet working and successful for everyone.
Adapted from original post which first appeared on TechDirt.
Austin Ruckstuhl is a Project and Policy Advisor at the Internet Society where he works on Internet impact assessments, defending encryption and supporting Community Networks as access solutions.
The views expressed by the authors of this blog are their own and do not necessarily reflect the views of APNIC. Please note a Code of Conduct applies to this blog.