January 6, 2025

The future of Internet censorship and Awala's role

By Gus Narea (Awala architect and Relaycorp CEO).

This blog post is adapted from the season finale of the Inside Awala podcast.

I want to share my thoughts on the emerging threats to Internet freedom, and the opportunities that the Internet Freedom community at large, and the Awala project in particular, will have to fend them off.

I have to be upfront with you: I believe that Internet freedom will continue to decline globally over the next 3–5 years before we get a chance to reverse this trend. This reversal is anything but guaranteed, and it’s going to take a multi-pronged approach involving technology, legal challenges, and advocacy efforts. I see Awala playing an increasingly important role on the technology side, but it’s not going to be enough on its own.

Let’s start by looking at what we know about Internet censorship today.

The State of Internet Censorship as of 2024

Nearly three-quarters of the world’s population lives in countries where Internet restrictions impair their human rights, and this situation continues to worsen. Freedom House reports that 2024 marks the 14th consecutive year of declining Internet freedom worldwide.

Last year broke several unfortunate records. For example, 167 major Internet shutdowns took place across 28 countries, which is the highest number of countries in a single year. Complete Internet blackouts lasted a total of 49,000 hours, twice as many as in 2023. Also, 648 million people were affected by deliberate outages. Overall, Internet censorship is estimated to have cost the global economy $7.69 billion in 2024.

The scale of investment in Internet censorship by repressive regimes is staggering. China alone spent an estimated $6.6 billion on its censorship apparatus in 2020, whilst Russia recently announced plans to invest a further $650 million over the next five years to enhance its capabilities. By contrast, the United States provided $90.5 million in the 2023 fiscal year for Internet freedom efforts globally. What’s more concerning is that the United States pretty much single-handedly funds the R&D to counter Internet censorship worldwide—indeed, we’re excessively reliant on a single country to advance the technical solutions to this global issue.

We are definitely underfunded, but I believe that we don’t need to match the budgets of repressive regimes dollar for dollar. Strategic investments, even modest ones, can help us turn the tide.

Internet Blackouts Will Become Less Common

Internet blackouts have been steadily increasing year after year when measured by their duration. Despite this trend, I predict that government-sponsored Internet blackouts will eventually become a thing of the past due to increasing reliance on the Internet, their massive economic impact, and, most importantly, the growing adoption of technologies and policies that will “splinter” the Internet instead. I think that the only exception to this will be North Korea for the foreseeable future.

Having said that, I do think that we’ll see a growing number of conflicts—like the wars in Gaza and Tigray—where adversaries will cut off civilians from the Internet. Additionally, climate change is likely to exacerbate conditions like resource scarcity, food insecurity, and mass displacement that historically lead to conflict.

In this context, Awala remains a crucial tool, but the stakes are significantly higher than with politically motivated blackouts. Whilst political censorship primarily threatens civil liberties, conflict zones put Awala users and couriers at direct physical risk as they attempt to maintain communications across battle lines or through dangerous territory. The consequences of discovery or interception could be catastrophic.

Splinternets Will Thrive

A splinternet is when a country deliberately isolates its portion of the Internet from the global network to force citizens onto domestic, state-controlled apps and websites. The extent of this isolation varies greatly—as of late 2024, China and Myanmar sit at one extreme with highly isolated networks, whilst Iceland stands at the opposite end with virtually no isolation. And even in China’s case, what we observe today is merely the tip of the iceberg: the Cyberspace Administration of China, which is responsible for the Great Firewall, continuously experiments with new isolation techniques, carefully calibrating how much control they can exert before triggering significant public pushback or economic consequences.

I believe Artificial Intelligence (AI) is going to accelerate this splintering of the Internet, as it will give governments powerful new tools to counter VPNs and other circumvention technologies, whilst also enabling them to create compelling domestic alternatives to Western services like YouTube and Facebook.

There are broadly two approaches to circumventing national firewalls like China’s Great Firewall:

• Using proxyless tools on your device, which essentially trick the firewall into allowing access to forbidden services. Whilst the firewall can still monitor the traffic, it lacks the capability to block it. This approach has proven particularly effective in Iran and other countries. However, I believe that within 3 to 5 years, AI-assisted software development, combined with modern hardware, will enable censors to rapidly identify and patch the vulnerabilities that these tools exploit. At that point, proxyless tools will likely only remain viable against the most basic national firewalls, and even those will eventually catch up.
• Using tools like VPNs and Tor, which act as middlemen between you and the services you want to access, routing your Internet traffic through servers to obscure which banned services you’re accessing from the firewall. Censors already work to discover the techniques used to hide this proxy traffic (known as “obfuscation techniques”), and to identify and enumerate the proxy servers themselves, but AI will dramatically accelerate both of these capabilities.

Using AI to identify the exploited holes will be particularly easy with open source circumvention tools. In fact, as part of my research into the state of the art in circumvention techniques whilst designing the Awala VPN last summer, I used AI to survey the obfuscation techniques employed by open source circumvention apps. You would think that, being open source, the technical details of their obfuscation techniques would be readily available, but they are not in many cases. Fortunately, I was able to get Google AI Studio to analyse their source code and identify the techniques used, taking less than two minutes per tool. Unfortunately, China, Russia, and others can do this too, and these days you do not even need proprietary American AI to do it; as I am writing this, DeepSeek V3, a new open source language model from China, is beating frontier models on coding benchmarks.

To be clear, I’m not advocating for closed source circumvention technologies—that would be short-sighted. Closed source tools simply offer the illusion of security, but the holes they exploit can still be observed from the outside, and AI can still expedite their discovery. If anything, I’m advocating for source-available tools and more transparency around obfuscation techniques. I designed the Awala VPN for this dystopian future, and, apart from making its source available, I also published its detailed technical design because I firmly believe in its ability to circumvent the national firewalls of the near future, but only by remaining extremely transparent about its obfuscation techniques.

Circling back to the server enumeration threat I mentioned earlier, I am deeply concerned that this enumeration capability may effectively spell the end of free proxy-based circumvention tools, unless we achieve a breakthrough that allows us to share server addresses exclusively with legitimate users whilst keeping them away from repressive regimes. I feel this is a field of research that isn’t getting the attention it deserves. To the best of my knowledge, only the Tor project is actively working on this, and they recently released a new system to distribute their server addresses, called “Rdsys”. Simply put, this system considers the reputation of a user requesting a server address—for example, the age of their Telegram account—and only shares an address with them if they’re deemed to have a good reputation. I believe this is a step in the right direction, but I’m not sure it’s going to be enough to counter the capabilities that censors will acquire in the coming years.

China offers a glimpse into a future where free circumvention services, including VPN services with free tiers, have largely been rendered ineffective. Today, Chinese residents must generally rely on premium VPN services or self-hosted VPN servers to reliably access the global Internet. Alternatively, many resort to so-called “airport services”, which are underground services operated within China that offer much faster speeds at a lower cost than foreign circumvention services. However, these domestic services lack basic privacy protections and may be subject to state surveillance.

Beyond blocking circumvention tools at the network level, repressive regimes will likely intensify their efforts to compromise devices themselves. From China’s mandatory spyware to North Korea’s controlled device supply chain, these regimes are increasingly focused on neutralising circumvention technologies before they even reach the network.

Having said all the above, these technical considerations about circumvention technologies may ultimately prove irrelevant in the face of a more fundamental challenge: the diminishing desire to access the global Internet in the first place. China’s approach demonstrates this vividly—they’ve built a parallel digital universe where domestic services are so polished and feature-rich that citizens have no reason to look elsewhere. The network effects are particularly powerful—with most Chinese speakers worldwide using these platforms, there’s little incentive to venture beyond them. Most concerningly, an entire generation of Chinese citizens has grown up within this walled garden, with no frame of reference for what lies beyond it. By making the domestic Internet ecosystem not just adequate but genuinely appealing, China has achieved what censorship alone never could.

What’s more concerning is that the emergence of region-specific generative AI could dramatically accelerate this trend around the world, enabling the creation of endless streams of localised, culturally-tailored content that keeps users thoroughly engaged within their domestic digital borders, making the allure of the global Internet even more distant.

Uncertainty in the Free World

Perhaps most troubling is how Internet freedom faces erosion from within the very democracies meant to champion it. The so-called “Free World” increasingly betrays its own principles, actively undermining the digital liberties it claims to protect.

Consider France and Spain, both members of the Freedom Online Coalition—a group dedicated to promoting Internet freedom—actively pursuing Internet censorship within their own borders. France recently blocked TikTok in New Caledonia, whilst Spain continues its efforts to block abortion-related websites. Meanwhile, the European Union pushes forward with legislation that would effectively undermine end-to-end encryption, a cornerstone of digital privacy and freedom. Most striking of all is India, the world’s largest democracy, which has earned the dubious distinction of being the global capital of Internet shutdowns, deploying this blunt instrument of digital control more frequently than any other nation.

I’m also watching with deep concern the spread of age verification legislation across numerous states in the US. This has already led major pornography websites to block access in those states. Predictably, residents are now turning to VPNs to circumvent this censorship—on American soil, no less. This raises a troubling question: how will these states respond? Will they attempt to implement state-level firewalls?

Equally concerning to me is the uncertainty surrounding a second Trump presidency’s impact on Internet freedom. The topic has been notably absent from campaign discussions and policy debates, making it impossible to anticipate the administration’s likely stance. Whilst I don’t anticipate major shifts in 2025, the subsequent years remain a complete unknown. (As a European, I really wish that this uncertainty will push the EU and the UK to put our money where our mouth is, and start providing adequate funding for Internet freedom efforts. This is, after all, a global issue.)

AI for Internet Freedom

AI represents not just a threat but also a tremendous opportunity for Internet freedom. Beyond the widely discussed productivity gains in software development, which I’m certainly experiencing firsthand, there are several other advantages we can leverage.

Take funding requests for Internet freedom initiatives which, in my experience, can take anywhere between five months and a year. AI could potentially expedite this process by helping requesters draft proposals, and helping both requesters and funders analyse proposals against funding parameters more efficiently. To do this, we will need local models like Llama that can run privately on laptops without external communication, which is crucial given the sensitivity of funding proposals. Having tried this myself recently, I don’t think that local models are good enough for this task yet, but I’m optimistic that this may change by the end of 2025.

Moreover, I believe that AI’s potential in outreach and marketing is particularly valuable for Internet freedom projects, especially in bridging cultural and linguistic barriers to reach diverse communities worldwide—at least for widely spoken languages. Communities using lesser-spoken languages may not benefit from this yet, unless somebody trains language models specifically for those languages.

As we navigate these challenging times, it’s clear that combating Internet censorship requires concerted effort from all of us. By staying informed and actively supporting initiatives that promote Internet freedom, we can contribute to reversing the current trend. I believe that with strategic investments and a multi-pronged approach, we can work towards a future where the Internet remains a place of free expression and access to information for everyone.