What does Europe’s new AI law mean for cities in Israel?
As of August 1, 2024, the European Union’s AI Act is in force. This isn’t just “European news” – over the next two years, its provisions will phase in and become the common language for working with vendors, cloud providers, and solutions that serve European residents. For municipalities in Israel, this means that procurement, security, and oversight of AI systems can no longer stay at the level of nice words – they now require procedures, documentation, and accountability.
Risk-based approach
The Act categorizes AI use cases into levels of risk:
- Prohibited – uses that are banned outright.
- High-risk – including education, hiring, essential services, critical infrastructure, transport, and biometrics – requiring strict oversight.
- Limited risk – with lighter requirements.
If a municipality operates candidate-screening systems, analyzes student performance, manages smart traffic lights, or deploys facial verification, it is already in high-risk territory. That brings obligations for risk management, data quality, documentation, logging, and explainability. The rule is simple: if it touches residents’ rights or safety, it ranks higher on the scale.
Key timeline
February 2025: Banned practices prohibited; AI literacy training required for staff and managers.
August 2025: Obligations for general-purpose AI (GPAI) models, including chatbots and generative tools, take effect – supported by a new code of practice on transparency and copyright.
August 2026: Broad application of the Act.
August 2027: End of transition period for existing GPAI models.
For Israeli municipalities, this means that questions on training data, copyright compliance, and testing procedures are no longer optional – they are part of every standard vendor conversation.
Practical impact on municipalities
Procurement: Add an “AI” section to tenders requiring conformity to the AI Act, disclosure of data sources, risk-management processes, a human-in-the-loop mechanism, and complaint/appeals procedures.
Ongoing projects: Create a municipal AI systems registry detailing system purpose, data sources, and responsible parties. For high-risk systems, conduct rights-and-privacy impact assessments, test for bias, and define edge cases before scaling citywide.
This may feel bureaucratic, but it prevents stoppages, fines, and builds public trust.
Examples on the ground
Transport: If upgrading traffic lights with AI, keep change logs, monitor performance against baselines, and ensure city policies (like public transport priority) remain embedded in the model.
Education: Before rolling out AI-driven assessment or behavior-monitoring tools in classrooms, clarify who sees reports, how to avoid over-labeling, and how to handle system errors involving students.
An opportunity, not just regulation
When the EU sets a standard, it helps municipalities align internally – across legal, procurement, digital, education, and transport departments – around a shared framework. Cities can already implement a “green lane” for AI projects: short supplier forms, targeted risk checks, and contract clauses ensuring access to logs, tests, and updates. Municipalities that start this now will be regulatory-ready by 2026, rather than scrambling to catch up.
