As a computer vision engineer, I have spent years teaching cameras to understand the world around a car. The same building blocks, cameras, image processing and machine learning, also power the surveillance systems that are spreading around the world. That makes the debate about “authoritarian tech” a very personal one for me.
A divide that keeps growing
In 2022, analysts described a new kind of cold war, fought not only with weapons but with technology. Several governments were investing heavily in camera networks, facial recognition and data platforms for policing, and exporting these systems abroad. At the same time, internet shutdowns were becoming a common tool to control protests and elections.
Four years later, the data shows the trend has continued:
- Internet freedom: Freedom House reports a 15th consecutive year of decline in global internet freedom, with conditions worsening in 28 of 72 countries in its 2025 report.
- Not only autocracies: half of the countries rated “Free” also saw their scores decline, including Germany and the United States.
- Record shutdowns: Access Now and the #KeepItOn coalition documented at least 313 internet shutdowns in 52 countries in 2025, the highest number since tracking began in 2016. Satellite internet was also blocked more often than before.
In other words, the line between “democratic” and “authoritarian” technology is not drawn by geography alone. It depends on how systems are designed, deployed and governed.
Regulation is starting to draw lines
Europe has taken a concrete step with the EU AI Act. Since February 2025, it has banned certain AI practices outright, including social scoring and, with narrow exceptions, real-time remote biometric identification in public spaces for law enforcement. Whatever one thinks of the details, the message is clear: some uses of technology are considered unacceptable, no matter how technically impressive they are.
What engineers can do
- Recognise dual use. Face detection, object tracking and data fusion are neutral techniques, but their deployment is not. Asking “who will use this, and for what?” belongs in every technical design review.
- Design for privacy. Data minimisation, on-device processing, anonymisation and strict retention limits reduce what can be misused later.
- Encrypt by default. End-to-end encryption and secure communication protect users even when infrastructure is under pressure.
- Build for resilience. Offline-first features and graceful degradation help people keep working and communicating when networks are disrupted.
- Know the rules. Regulations like the EU AI Act are becoming part of the engineering requirements, just like functional safety standards in automotive.
The takeaway
Technology does not choose sides by itself; people and institutions do. As engineers, we cannot control every way our work is used, but we can make privacy, security and transparency part of the design rather than an afterthought.
Sources
- MIT Technology Review, “The world is moving closer to a new cold war fought with authoritarian tech”, September 2022. Link
- Freedom House, “Freedom on the Net 2025: An Uncertain Future for the Global Internet”. Link
- Access Now, “Rising repression meets global resistance: Internet shutdowns in 2025”, March 2026. Link
- Lewis Silkin, “Prohibited practices and AI literacy requirements of EU AI Act now applicable”, February 2025. Link