Friday Digest #6: The Cost of Internet Shutdowns in 2023

Hi, it’s Sam. Welcome to our first newsletter of 2024, I hope you had a relaxing break!

Earlier this week, we published our annual Cost of Internet Shutdowns report for the fifth year in a row. It shows that deliberate, self-imposed internet shutdowns affected almost 750 million people and cost the global economy $9.01 billion (USD) in 2023.

In this edition, we’ll take a quick look through the main findings and discuss some of the challenges we confronted while working on the report this year.

Cost of Internet Shutdowns 2023 Findings

Throughout 2023 we recorded 196 major self-imposed internet shutdowns in 25 countries, with an estimated cost of more than $9 billion (USD). Russia was by far the worst affected again this year, with losses amounting to $4 billion, followed by Ethiopia ($1.6 billion) and Iran ($920 million).

Like 2022, X (Twitter) was the most affected platform last year, suffering more than 10,000 hours of disruption worldwide, 18% more than Instagram and nearly 30% more than TikTok.

For the first time since we began tracking shutdowns, Iraq emerged as the country that blocked access the most frequently, with 66 shutdowns in 2023. Almost all of these occurred during school exams in a supposed bid to prevent cheating.

Meanwhile, the longest shutdown beginning in 2023 occurred in Manipur, India. The restrictions, which lasted over 5,000 hours, cost an estimated $125 million.

As ever, there was a wide variety of reasons internet shutdowns occurred and while exams prompted the most restrictions, conflict-related internet restrictions were by far the most costly.

A Note on Our Methodology

Estimating the financial impact of internet disruptions is not easy. Just identifying when, where and why internet shutdowns have occurred requires constant monitoring of a huge variety of data sources. On top of that, there’s a number of variables at play as every country has unique technical and economic characteristics.

To provide a consistent estimate we use Netblocks’ COST tool, which is based on a method originally devised by the Brookings Institution, with CIPESA’s specialized model used for sub-Saharan Africa. It uses various economic indicators from the World Bank to the International Telecom Union.

But the tool only offers nationwide estimates and often, particularly in a country like India, internet shutdowns are regional and only affect a portion of society. To calculate these, we use the region’s economic output as a proportion of national GDP to modify the COST tools’ results. Meanwhile, partial internet outages are calculated as a proportion of the above costs based on the most up-to-date internet market-share information publicly available for the affected country.

There are also instances when the financial implications of internet restrictions are almost impossible to calculate and potentially misleading. After much deliberation, we decided to exclude recent restrictions in both Ukraine and Gaza because they weren’t self-imposed — that is, the restrictions were caused by a foreign actor. Moreover, the economy is already so heavily affected by conflict that the financial implications of the restrictions become nearly impossible to decipher.

Finally, there’s also an issue implicit within the definition of an internet shutdown: namely, at what point does a disruption stop being a shutdown and become ongoing internet censorship? There aren’t many people, for example, that describe China’s blocking of Western social media platforms as an internet shutdown.

Both Russia and Iran imposed major social media restrictions in 2022 that showed no sign of ending in 2023. After careful consideration, we took the view that the economic cost of social media restrictions should be included in our calculations for up to 365 days. After this point, we exclude them as the restrictions represent ongoing internet censorship and many of the economic consequences will likely have been overcome by people switching to other platforms.

As we continue to work on this into 2024, I’m sure we’ll encounter new challenges and I’m looking forward to building on this methodology over the next 12 months.

If you would like to discuss the findings or methodology in more detail please contact us: research@top10vpn.com

What We’ve Been Reading

• Washington Post: How the battle for democracy will be fought — and won

  This important op-ed from Washington Post locates internet censorship and censorship circumvention technologies in the global struggle for democracy. It highlights innovative methods to bypass censorship and promote free expression like Toosheh, a method used to deliver uncensored information into Iran via satellite, bypassing internet blockades.

• BBC: Google settles $5bn lawsuit for 'private mode' tracking

  Google has agreed to settle a $5 billion lawsuit in the U.S. alleging that it invaded user privacy by tracking them in "private mode." The lawsuit, filed in 2020, accused Google of collecting data even when users set their browsers to private or incognito modes, a practice the company defended as necessary for evaluating site performance.

• TechCrunch: 23andMe tells victims it’s their fault that their data was breached

  23andMe is facing over 30 lawsuits due to a massive data breach affecting 6.9 million users. But in response, the company is blaming the victims, claiming their negligence in password security led to the breach. The breach was a result of hackers accessing accounts through credential stuffing and exploiting the DNA Relatives feature, compromising data of users who did not directly use recycled passwords.

• Hao-Ping (Hank) Lee et. al., Deepfakes, Phrenology, Surveillance, and More! A Taxonomy of AI Privacy Risks

  This paper analyzes 321 AI privacy incidents and constructs a useful taxonomy of AI privacy risks, identifying 12 high-level risks that AI technologies either newly created or exacerbated. It shows that incorporating AI into products changes the privacy risks involved, but current privacy-preserving methods in AI/ML, such as federated learning and differential privacy, only address a subset of these risks.

• Gizmodo: Meet ‘Link History,’ Facebook’s New Way to Track the Websites You Visit

  Facebook has introduced a new "Link History" setting in its mobile app, which tracks and stores all links clicked by the user for targeted advertising purposes, though users can opt out. The setting is part of Meta's broader strategy to maintain its data harvesting practices amid increasing privacy restrictions from tech companies and lawmakers, even as it raises questions about the extent of privacy protection it offers.