AI is watching you: The global spread of surveillance

Hey, I'm Michelle, the curator of Datawrapper’s weekly Data Vis Dispatch. This week, I took a break from my usual tasks of collecting and finding interesting data vis to write my first-ever Weekly Chart. (Definitely check out this week’s collection from Lisa!)

A few months ago, I read that Germany’s Federal Office for the Protection of the Constitution had decided to drop Palantir, a U.S. data analytics and defence software company, from its federal security contracts. Germany is just one of many countries that have been reported to be deploying various AI-powered surveillance technologies to create massive data profiles using artificial intelligence and predictive analytics.

Surveillance itself is nothing new, but AI has changed its scale. Where surveillance once took place manually by watching camera footage or going through it by hand, AI systems can now do this automatically. They recognise faces in crowds, analyse movement patterns, and combine data from cameras, phones, social media and government databases into detailed profiles.

Governments usually justify these tools with public safety, crime prevention or counter-terrorism. Critics, however, warn about privacy violations, errors and bias, such as facial recognition misidentifying people, and the risk of threatening core democratic principles.

For my Weekly Chart, I used data from the AI Global Surveillance (AIGS) Index, a research project by the Carnegie Endowment for International Peace that tracks the spread of AI and big data surveillance tools. Spanning the years 2012 to 2022, the index covers 97 countries, where open-source analysis indicates the adoption of AI technology. It’s important to note that the index does not distinguish between legitimate and unlawful uses of AI surveillance, but instead documents the spread of surveillance capabilities and identifies the technologies and companies involved.

What technologies are being deployed?

The AIGS Index groups surveillance technologies into four categories:

• Safe city platforms: City-wide networks of cameras and sensors, often linked to a central command centre

• Facial recognition systems: Cameras that identify people by matching their faces against databases

• Smart policing: Data-driven tools, such as predictive policing, that analyse large amounts of data to anticipate crime or locate suspects

• Social media surveillance: Monitoring of social media platforms and online activity

The use of surveillance has spread across all political regimes. In fact, and maybe surprisingly, liberal democracies have the highest usage rates for two of the four technologies: 95% use facial recognition and 91% use smart policing. Autocracies, by contrast, rely more heavily on social media surveillance (62% of closed autocracies, compared with 41% of liberal democracies). The difference between regime types therefore lies in which tools they choose and how they use them.

Surveillence suppliers: United States vs. China

The index also records which companies and countries supply these technologies. Governments choose a supplier for many reasons, including legal, ethical, geopolitical, and security considerations. The supplier's country of origin matters particularly where countries differ in their regulatory frameworks, government oversight, data protection rules, or geopolitical relationships.

This is where two further indices come in. The Digital Repression Index (DRI) and the V-Dem Electoral Democracy Index (EDI) are two complementary frameworks that help us analyse the relationship between technology and governance.

While the EDI focuses on electoral aspects of democracy, the DRI measures how much a government uses digital tools to repress its population, for example through online censorship, internet shutdowns, or arrests of internet users. Together, they track suppliers of surveillance technology and their clients among governments.

The need for stronger AI surveillance regulations

With so many governments deploying automated surveillance systems, using these tools responsibly is becoming increasingly important. Yet regulation has not kept pace with the technology.

Around the world, laws intended to restrict AI surveillance and ensure its proper use are either incomplete, being postponed or, in some cases, absent. This creates gaps in oversight, leaving people more exposed to potential misuse. While recent regulations, such as the EU's AI Act, are a step in the right direction, concerns remain that they are insufficient. Stronger, more comprehensive legal frameworks are especially needed to keep up with the rapid spread of these automated technologies.

That’s all from me! Thanks for reading. I look forward to seeing you in another Data Vis Dispatch next week, while our co-CEO Elana will be back with the next Weekly Chart.