Shutdown Tracker Optimization Project (STOP) 

The #KeepItOn Shutdown Tracker Optimization Project (STOP) is created and maintained by Access Now, as part of the #KeepItOn coalition’s global efforts to fight internet shutdowns. STOP aims to document and contextualize internet shutdown cases around the world, under the definition of internet shutdowns developed by the #KeepItOn coalition. Contact #KeepItOn Data Analyst Zach Rosson ([email protected]) and #KeepItOn Campaign Manager Felicia Anthonio ([email protected]) with any questions or comments.


a note on our data

To date, the tracker contains data from 2016 through 2023. While we try to build a comprehensive dataset, we rely on technical measurement as well as contextual information, such as news reports or personal accounts. The constraints of our methodology mean that there may be cases of internet shutdowns that have gone unreported, and numbers are likely to change if and when new information becomes available.

About our methodology

When STOP was launched in 2017, it was a first-of-its-kind project systematically documenting internet shutdowns worldwide. Our goal is to paint a comprehensive picture of the shutdown landscape, which complements technical data with vital contextual information on how shutdowns are impacting people and communities at risk, provided by local civil society actors, human rights defenders, journalists, and other partners on the ground.

As such, we chose from the outset to combine quantitative and qualitative data: we used first-hand accounts, news articles, official government orders, internet traffic measurements, and more to determine exactly what happened during each shutdown. Combined with the results available from technical measurement tools at the time, this allowed us to identify and map a range of shutdown instances, from neighborhood-specific shutdowns all the way up to countrywide blackouts. We also tracked government statements and claims of attribution – or more often, denials and silence – in hopes of clarifying responsibility and holding shutdown perpetrators accountable.

Since 2017, the shutdown measurement community has greatly expanded its capacity for monitoring, identifying, verifying, and documenting shutdowns, through technical achievements and rigorous research. Granular, country-specific datasets from #KeepItOn coalition partners, along with ongoing collaboration, reporting, and resources from internet measurement groups, have helped grow and improve shutdown research, datasets, and advocacy. The increasing quality of STOP is a testament to our partners’ ongoing support, and the wider efforts of the #KeepItOn coalition and community as a whole.

Ultimately, our role at Access Now is to compile all of this information, manually verify the details from a variety of sources, and present it in a way that facilitates and extends our partners’  research and advocacy work. In the process, we refined and honed how we structure and report shutdown data. While any errors or inaccuracies are ours alone, we are committed to representing and amplifying the #KeepItOn coalition and community’s efforts, meeting our partners’ needs, and expanding the evidence base for ending internet shutdowns.

Based on feedback and questions from stakeholders, we refreshed our methodology in 2024 to address key issues and explain our reasoning, including the scope of our approach.

Frequently asked questions

How do you define an internet shutdown? What is a shutdown instance?

An internet shutdown is an intentional disruption of internet or electronic communications, rendering them inaccessible or effectively unusable, for a specific population or within a location, often to exert control over the flow of information.

STOP was designed as a contextual dataset. We record shutdown instances to identify and attribute shutdowns to particular actors, document the specific human rights harms of shutdowns, support efforts to establish a coordinated action plan to enable alternative access to internet services, and to hold the perpetrators of shutdowns accountable.

A shutdown instance refers to any disruption event lasting longer than one hour, or a series of disruption events that we attribute to the same or similar circumstances, justifications, methods, and perpetrators. Depending on the circumstances, such as the status of local laws, the ability of each actor involved, and the state of humanitarian crises, a shutdown instance may result from one or several different technical events, which may change in scope during the instance in question.

What types of shutdown instances do we document in the tracker?

Shutdowns are deliberate actions taken by authorities to silence, isolate, or restrain people during protests, examinations, conflict, elections, and other important events. They continue to grow in number and technical complexity, with many “repeat offenders” hitting the kill switch year after year. As a result, shutdowns can come in many forms, and our tracker looks at full network shutdowns, bandwidth throttling, service-based blocking for two-way communication platforms, and sometimes a combination of all three across one shutdown instance. We also determine whether mobile and/or broadband networks are affected, and the geographic extent of the shutdown. (All of our column indicators can be reviewed below in Metadata).

Apart from the one-hour minimum requirement, we only consider disruptions that we can attribute to a specific cause. Through a combination of multiple quantitative and qualitative sources, we carefully verify recorded internet disruptions and blockings, and attribute them to specific actors as deliberate internet shutdowns. Imposing parties include all levels of government, the judiciary, militaries, and police forces, some of which occasionally implement shutdowns across borders and in occupied territories. Our tracker includes shutdowns imposed by deliberate disconnection or interference at internet service providers (ISPs), as well as shutdowns caused by the targeting of civilian infrastructure through airstrikes, fuel embargoes causing infrastructure failure, cyberattacks, and more.

Why is a platform block considered an internet shutdown? How does this differ from censorship of websites?

For billions of people worldwide, online platforms with peer-to-peer messaging and information-sharing functions are the main way of accessing and using the internet, allowing efficient and immediate communication, access to information, and (ideally) free expression. They play a vital role in enabling human rights, whether during disaster response, coordination of protest movements, elections, or accessing humanitarian corridors during conflict. As a result, based on our definition, any interference with these services intended to render them inaccessible or effectively unusable is considered a contravention of human rights and an internet shutdown.

For the purposes of an internet shutdown, peer-to-peer messaging and information sharing can be broadly defined as two-way or multi-way exchange of information through platforms. In comparison, broadcast media and news publishing as accessed through the internet are effectively one-way communication. Deliberate blockings of these sites are clear acts of censorship, and interfere with the exercise of freedom of expression and access to information, but fall outside the specific scope of an internet shutdown for advocacy purposes.

Why are ongoing platform blocks counted the same way as individual shutdown instances? Why are some cases missing from your dataset?

With governments increasingly turning to targeted forms of shutdowns, whether focused on specific geographies or mobile networks, platform blocks have become a tool to quash dissent and control the flow of information. Alongside broader censorship of news and other websites, governments often deliberately block platforms for months or even years at a time. Similar to other ongoing shutdowns extending over multiple calendar years, we count ongoing platform blocks once per every year in which they occur, to account for the ongoing, deliberate decision of leaders to cut off access to a platform.

Since 2019, we’ve tracked shutdowns of Facebook, X (formerly Twitter), WhatsApp, Instagram, and Telegram in five separate columns of our dataset as they are some of the most frequently and widely used platforms. However, we include other service blocks, as they arise, in the other_affected column (e.g. Signal, TikTok, Grindr), and continue working to include all impacted platforms as they become more popular and thus prone to targeting by authorities. (See descriptions of all columns in our dataset in the Metadata section below).

Overall, platform blocks have historically been under-reported, and our team has engaged in an ongoing process of updating the STOP dataset to include more of these and other longstanding but previously under-counted shutdowns. Getting an accurate picture of the full scope of platform blocks underway around the world is a challenge due to the localized nature of platform preferences, limited measurement data for platforms in countries where the user base is relatively small, discrepancies when a government does not fully or consistently implement stated policies for platform blocking, and people’s regular use of VPNs and other similar tools to circumvent blocks. To correct this, Access Now is adding missing data on longstanding platform blocks, and will continue to work closely with #KeepItOn coalition partners, including the Open Observatory of Network Interference (OONI), to strengthen our documentation in this area and hold all perpetrators to account.

What metrics do we use to track and report on shutdowns? Why do we count shutdowns differently depending on context?

In addition to documenting the type, scope, cause, and justification of each shutdown in our dataset, we also track metrics including duration, correlation with violence and serious human rights abuses, and specific groups targeted. This data allows us to analyze trends over time and identify new and ongoing impacts of shutdowns. (Descriptions of all columns are available in the Metadata section below.) In the course of our reporting on shutdowns around the world, we count shutdowns by country and region to show where shutdowns are happening and roughly how often to hold perpetrators to account and direct attention to impacted areas. These counts are broadly a collection of the total number of times authorities decided to shut off the internet for people within a given location.

In every shutdown instance, a government, military, or other ruling power ultimately makes the choice to cut off the internet. Whether it’s through a multi-day court order, a series of shutdowns rolled out during school exam periods, or intermittent shutdowns deployed during ongoing protests, a shutdown instance most often results from one consequential decision made by someone in a position of authority. Based on contextual information on how and why a shutdown was implemented, we define a shutdown instance and collect all the relevant details linked to the single action by an authority figure. This means a government that makes numerous snap decisions over the course of ongoing protests would end up being linked to numerous shutdown instances, while a predetermined, single policy of shutdowns during a school exam period would end up being linked to one shutdown instance. These factors, as well as the discrepancy in data coverage in many regions around the world, result in slight variations when determining a shutdown instance and counting shutdowns across different contexts.

Why are months or years-long shutdowns counted the same way as shutdowns lasting just a few hours or days? How can we better reflect the severity of shutdowns in our data?

The duration of shutdowns can vary widely, from months to years during protracted conflicts, to hours-long shutdowns during protests. Whether short or long, however, they can have severe impacts, covering up police brutality and human rights abuses during shorter, targeted disruptions, or shrouding atrocities and war crimes during prolonged shutdowns. Our aim is to call attention to the harms of every shutdown, no matter how short or long, so that no one endures a shutdown in silence. As described in the previous question, (What metrics do we use to track and report on shutdowns? Why do we count shutdowns differently depending on context?), shutdown instances are determined by their linkage to authorities’ actions. Other than a minimum duration of one hour, shutdown instances can be defined at any length. As a result, with our current approach, total counts of shutdown instances by country group together shutdowns of all durations.

The metrics described in the previous question – total counts by country, scale, duration, correlation with serious human rights abuses and violence, and groups targeted – can all be used to emphasize the severity of shutdowns. However, in addition to quantitative measures, the #KeepItOn Shutdown Impact Stories project continues to highlight the wide-ranging experiences of people around the world living with the impacts of internet shutdowns. Their stories have spurred unequivocal condemnation and are key evidence for documenting the human harms of internet shutdowns. Our annual #KeepItOn report also provides contextual information and local nuance, highlighting the people affected and specific harms which make clear that shutdowns are incompatible with human rights.

However, we do wish to acknowledge that, while our approach to counting shutdowns works well in many regions and contexts, in others it can serve as a detriment to our partners’ goals or misrepresent the complexity of how shutdowns are happening. Whether inadvertently or on purpose, journalists and media organizations can misrepresent, distort, or simplify STOP data to fit certain narratives or compare countries. And state actors can justify or minimize their own actions, while purposefully ignoring or omitting facts for their own political gain.

For example, in the Asia Pacific region where authorities in multiple countries are some of the most entrenched shutdown perpetrators, India has consistently led the world with the highest total number of recorded shutdowns each year, documentation of which has been possible in part thanks to court orders requiring authorities to publish shutdown orders. It is important to note that other countries in the region also have alarming practices regarding shutdowns, but documentation of these may be less clear. In Myanmar, for example, reports indicate that the military junta has likely implemented hundreds of localized shutdowns during its operations in recent years, which have gone undocumented. The number of shutdowns recorded in STOP for Myanmar since 2021 leaves many of those likely shutdowns uncaptured, as independent verification has become increasingly difficult amid crackdowns on digital rights activists, decentralized decision making, and continued degradation of infrastructure. Although #KeepItOn coalition partners have worked tirelessly to uncover evidence of these shutdowns, we continue working actively to ensure the data presented through STOP more accurately reflects the severity of shutdowns happening across Myanmar since the coup.

Systematically documenting and reporting on human rights abuses will always be an imperfect process. We are nevertheless committed to improving how we represent shutdown severity, exploring new avenues and metrics for understanding risks and harms, and representing regional differences in shutdown tracking to the best of our ability.

Why are there so few recorded shutdowns from Russia and China? What about North Korea?

A number of factors contribute to historically and comparatively lower numbers of reported shutdowns in the Russian Federation and People’s Republic of China. In both countries, systematized censorship, surveillance, and legally entrenched use of a domestic intranet significantly decrease authorities’ need to use shutdowns as a tool to stifle dissent or control information flows. In parallel, activists and technologists in these countries face insurmountable threats and challenges when sharing shutdown reports and data. In previous years, both countries have nonetheless imposed targeted shutdowns in various contexts, as well as ongoing platform blocks as part of persistent crackdowns on dissent.

Since its illegal invasion of Ukraine in 2022, the Russian military has imposed a barrage of shutdowns, underscoring the serious harms that Russia’s government has inflicted beyond its borders. Russia has also blocked multiple social media platforms in recent years across the country, including Facebook, Instagram, and X (formerly Twitter). So far in 2024, reports indicate that Telegram and WhatsApp were throttled across multiple regions, and there are separate reports of YouTube also being throttled.

In China, many ongoing shutdowns went into effect before we started collecting STOP data in 2016, with some dating back to 2009. The current dataset captures more recent shutdowns, including blocks of Signal and Grindr ongoing in the country since 2021. Meanwhile, a 2023 incident damaging undersea cables in the Taiwan Strait, which caused an internet disruption on a Taiwanese island, raised concerns that the Chinese government was attempting to extend its control over internet access, communication channels, and information flows.

In North Korea, the global internet is generally only available to a small group of wealthy elites and academic institutions, with everyone else only able to use the state-run, “walled garden” intranet, Kwangmyong. This level of authoritarianism and isolation has historically prevented the North Korean government from needing to resort to internet shutdowns — but the country reportedly still blocked Facebook, X (formerly Twitter), and YouTube in 2016. With limited ability to monitor internet traffic and censorship in North Korea, identifying new shutdowns and determining if platform blocks are ongoing remains an active area of work.

Data sources

Our data is sourced from the #KeepItOn coalition, the measurement community, news and media outlets, and other channels. Below are some databases and resources we use for research, data collection, and verification.

SourceLink
Internet Outage Detection and Analysis (IODA)https://ioda.live/
Cloudflare Radarhttps://radar.cloudflare.com/
Open Observatory of Network Interference (OONI)https://ooni.org/
Software Freedom Law Center India (SFLC.in)http://internetshutdowns.in/
Athanhttps://athanmyanmar.org/
Myanmar Internet Projecthttps://www.myanmarinternet.info/
Filter.watch (Miaan Group)https://filter.watch/
Digitally Righthttps://digitallyright.org/
Yodethttps://shutdowns.yodet.org/
Kill Switch in Pakistanhttps://killswitch.pk/
Internet Society Pulsehttps://pulse.internetsociety.org/
Réseaux IP Européens (RIPE) Atlashttps://atlas.ripe.net/
Censored Planethttps://censoredplanet.org/
Measurement Lab (M-Lab)https://www.measurementlab.net/
Google Transparency Reporthttps://transparencyreport.google.com/traffic/overview/
Mozilla Telemetry Datahttps://telemetry.mozilla.org/
Government of Bihar, Home Departmenthttps://home.bihar.gov.in/
Government of Jammu & Kashmir, Home Departmenthttps://www.jkhome.nic.in/orders.html
Government of Haryana, Department of Home Affairshttps://homeharyana.gov.in/
Government of Manipur, Home Departmenthttps://manipur.gov.in/
Government of West Bengal, Home & Hill Affairs Departmenthttps://home.wb.gov.in/
Meta Transparency Report (until 2022)https://transparency.meta.com/reports/internet-disruptions/
Oracle Internet Intel (2018)No active link
Moses Karanja (2016, 2017)https://moseskaranja.com/tracking-internet-shutdowns-in-africa/
Collaboration on International ICT Policy for East and Southern Africa (CIPESA) (2016, 2017)https://cipesa.org/

Metadata

Column nameColumn description

start_date_type
Categorical data describing if the start date of the shutdown is known through measurement data or other sources, or is the best estimate based on the available information.– Actual
– Estimated
start_dateThe date during which the shutdown began.MM/DD/YYYY format
countryThe country where the shutdown occurred. In most cases, the government of this country is also the perpetrator of the shutdown. However, when the shutdown was imposed by a third-party or outside military, this column indicates the country affected by the shutdown while ordered_by and decision_maker gives the details on the actual perpetrator.Follows ISO 3166 naming convention
geo-scopeCategorical data describing the geographic scope of the shutdown.– It only affected one city, county, or village (Level 1)
– It affected more than one city in the same state, province, or region (Level 2)
– It affected locations in more than one state, province, or region (Level 3)
Not sure
area_nameFurther specifies the geographic area impacted by the shutdown. Ranges from specific neighborhoods to multiple regions depending on the shutdown. If multiple areas are affected, we list all relevant locations.E.g. Nationwide
shutdown_typeCategorical data describing if the internet was cut off, intentionally slowed down, or a combination of the two during the shutdown instance.– Shutdown
– Throttle
– Shutdown, Throttle
affected_networkCategorical data describing the internet access points affected by the shutdown.– Broadband
– Mobile
– Broadband, Mobile
– Unknown
shutdown_extentCategorical data describing if the whole internet was impacted through that access point, if a specific service or platform was impacted, or if it was a combination of the two during the shutdown instance.– Full network
– Service-based
– Full network, Service-based
ordered_byCategorical data describing which entity ordered the shutdown. Most often, this is an authority within the country above, but can also include an outside actor described further in decision_maker.– Executive government
– Judiciary
– Law enforcement
– Local government
– Military
– Non-government
– State government
– Unknown
decision_makerFurther specifies which person, government entity, or military chose to shut down the internet.E.g. Government of Rajasthan, Home Department
E.g. Military of the Russian Federation
actual_causeCategorical data describing the primary trigger for the shutdown. When there are multiple triggers and additional context, the event column will give further detail.– Communal violence
– Economic crisis
– Elections
– Exam cheating
– Information control
– Other
– Political instability
– Protests
– Religious holiday/anniversary
– Unknown
Visits by government officials
actual_cause_detailsFurther specifies what triggered the shutdown.E.g. Missile strikes intentionally targeting energy infrastructure
info_source
Categorical data describing the primary source of evidence of the shutdown.
– Confidential
– CSO KIO partners
– News media article
– Other
– Social media
info_source_linkReference URL of the primary source plus additional URLs used as evidence.List of sources separated by a semicolon
shutdown_statusCategorical data describing if the shutdown has ended, is ongoing, or if the status is unknown based on the available information at the end of each calendar year. – Ended
– Ongoing
– Unknown
end_dateThe date during which the shutdown ended.MM/DD/YYYY format
Blank if shutdown_status
Ended or Ongoing
durationThe number of days between start_date and end_date, including the first day.If a shutdown starts and ends on the same day, no matter the duration in hours, duration = 1.
gov_justificationCategorical data describing the government’s stated justification for imposing the shutdown when the information is available.– Fake news/ Hate speech/ – Incendiary content/ Promoting violence
– Illegal content
– National security/ Counter-terrorism
– None
– Other
– Precautionary measure
– Public safety/ Quell unrest/ Restore public order
– Sabotage/ Third-party Action
– School exams
– Technical problems/ Infrastructure failure
– Unknown
gov_just_detailsFurther specifies how the government justified the shutdown based on gov_justification.Provides context for gov_ack_quote
gov_ackBoolean data describing if the government acknowledged the shutdown in any manner of public statement or quote by a government official.– Yes
– No
– Unknown
legal_justBoolean data describing if the government or relevant authority gave any legal justification or referred to a specific law.– Yes
– No
– Unknown
legal_methodFurther specifies the memo, court order, or similar instrument used as part of legal_just.E.g. Section 69A of the Information Technology Act, 2000
gov_ack_sourceURL source of the government order or news media article giving an official statement or quote, or otherwise proving acknowledgment by the perpetrator.List of sources separated by a semi-colon
gov_ack_quoteThe quote pulled from gov_ack_source.Depending on the source and context, may summarize the perpetrator’s acknowledgment
facebook_affectedBoolean data describing if Facebook has been blocked or throttled during the shutdown instance. If shutdown_extent is “Full network”, then the result will be “No”.– Yes
– No
– Unknown

twitter_affected
Boolean data describing if X (formerly Twitter) has been blocked or throttled during the shutdown instance. If shutdown_extent is “Full network”, then the result will be “No”.– Yes
– No
– Unknown
whatsapp_affectedBoolean data describing if WhatsApp has been blocked or throttled during the shutdown instance. If shutdown_extent is “Full network”, then the result will be “No”.– Yes
– No
– Unknown
instagram_affectedBoolean data describing if Instagram has been blocked or throttled during the shutdown instance. If shutdown_extent is “Full network”, then the result will be “No”.– Yes
– No
– Unknown
telegram_affectedBoolean data describing if Telegram has been blocked or throttled during the shutdown instance. If shutdown_extent is “Full network”, then the result will be “No”.– Yes
– No
– Unknown
other_affectedFurther specifies additional platforms blocked, if applicable, or if the platform or service affected none of the above five (Facebook, X/Twitter, WhatsApp, Instagram, Telegram).E.g. TikTok
E.g. Grindr
E.g. Signal
sms_affectedBoolean data describing if SMS/text messaging services are also impacted alongside the internet shutdown.– Yes
– No
– Unknown
phonecall_affectedBoolean data describing if phone services are also impacted alongside the internet shutdown.– Yes
– No
– Unknown
telcos_involvedThe telco/internet service provider (ISP) involved in the shutdown. In the majority of cases, these companies were given a shutdown order by a government entity. In some cases, particularly during conflict when outside militaries impose shutdowns, the telco listed had its services impacted as determined through traffic data visible by Autonomous System Number (ASN) or other sources.E.g. Jordan: Zain
E.g. Palestine: PalTel
telco_ackBoolean data which specifies whether the ISP(s) listed in telcos_involved acknowledged the shutdown.– Yes
– No
– Unknown
telco_ack_sourceURL source of a statement or quote by an ISP acknowledging the shutdown.List of sources separated by a semi-colon
telco_ack_quoteThe quote pulled from telco_ack_source.Depending on the source and context, may summarize the ISP’s acknowledgment
electionBoolean data describing whether the shutdown was triggered by an election.– Yes
– No
– Unknown
violenceBoolean data describing whether there was any reported violence during the shutdown.– Yes
– No
– Unknown
hr_abuse_reportedBoolean data describing whether grave human rights abuses were documented during the shutdown. This includes evidence of violence, including murder, torture, rape, or apparent war crimes by governments, militaries, and police or security forces.– Yes
– No
– Unknown
users_targetedThe specific groups impacted by the internet shutdown, if any.E.g. LGBTQ+ people
E.g. Ethnic minorities
users_target_detailFurther specifies the groups affected in users_targeted.E.g. Baluchi ethnic minority
users_notifiedBoolean data describing whether the public received advance warning of the shutdown.– Yes
– No
– Unknown
econ_impactThe economic impact, typically in U.S. dollars, of the shutdown according to estimates from sources in info_source_link.This column isn’t populated after 2018 in STOP
eventA brief summary of the shutdown and its context, including time of day when available and the types of impacts.Provides the most concise summary of the shutdown instance
an_linkURL source of posts, statements, publications, or press releases from Access Now about the shutdown.List of sources separated by a semicolon
regionCategorical data describing the region where the shutdown occurred.– Africa
– APAC (Asia-Pacific)
– EECA (Eastern Europe and Central Asia)
– Europe
– LAC (Latin America and the Caribbean)
– MENA (Middle East and North Africa)