Organisations in Israel
AI and the Privacy Protection Law in Israel
Israel has no AI law, so the Protection of Privacy Law, as amended by Amendment 13 from 14 August 2025, is the main binding rule for AI that processes personal data. The Privacy Protection Authority's draft guidance on AI systems, published for consultation in 2025, says the law applies across the AI lifecycle. It expects informed consent that explains how the AI system processes data, privacy notices that mention AI, a data protection impact assessment before processing personal data with AI, a policy on staff use of generative AI, accurate outputs, and strict handling of access and correction requests. It treats scraping personal data to train AI without consent as unlawful.
Israeli organisations govern AI mainly through privacy law. The government’s AI policy chose not to legislate for AI as a whole, so the binding rules come from existing law applied by sector regulators. For any AI that touches personal data, the most important of those is the Protection of Privacy Law, which Amendment 13 overhauled from 14 August 2025. This guide sets out how the Privacy Protection Authority (PPA) applies it to AI, as of October 2026.
Why privacy law is the starting point
Amendment 13 broadened the law in ways that reach most AI systems. Personal information now covers any data about an identified or identifiable person, including online identifiers. Information of special sensitivity includes biometric data, location data and personality assessments, which are the kinds of data many AI systems use or infer. Notice duties are wider, many organisations must appoint a privacy protection officer, and the PPA can impose administrative fines.
The law has no rule aimed at automated decisions. Its duties apply to AI through consent, notice, purpose, accuracy, security and access rights, and the PPA’s draft guidance explains how.
What the PPA’s draft guidance says
The PPA published draft guidance on applying the Protection of Privacy Law to AI systems for public consultation in 2025. It covers the full lifecycle, from collecting and preparing training data to testing, deployment and use.
Informed consent. Where consent is the basis, people need to understand what they are agreeing to. The draft expects notice to cover the purposes, the recipients, whether providing the data is mandatory, the types and sources of data, the risks, how the AI system processes the data, and whether an AI system is collecting it where that could materially affect the decision to consent.
Transparency. Privacy notices should expressly mention the use of AI. Organisations should consider telling people when they are dealing with an AI system, such as a chatbot.
Scraping. Scraping personal data from the internet to train AI without consent is unlawful in the PPA’s view, and the fact that a post is public does not mean the person consented. Operators of platforms are expected to take measures against scraping, and unauthorised scraping may amount to a serious security incident that has to be reported.
Impact assessment. The PPA recommends a data protection impact assessment before processing personal data with AI, to identify privacy risks, consider less intrusive alternatives and set mitigations.
Accuracy. Organisations should take care over the accuracy of data AI systems produce. Correcting an inaccurate output may require changes to the system itself, not only to a record.
Rights. Access, correction and deletion requests will be enforced for AI systems as for any other database.
Generative AI policy. Organisations should adopt a policy on staff use of generative AI tools, covering who may use them, which tools are allowed, what data may be entered, how long prompts are kept, whether inputs may be used for training, and training for employees on the risks.
Putting it into practice
The draft guidance turns into a short set of controls.
- Know what AI is in use. Many AI tools arrive inside existing software or are adopted by staff directly. A shadow AI review finds them, and an AI register records each one with the personal data it touches.
- Check the basis and the notice. For each system, confirm the basis for processing and that privacy notices describe the AI use in the terms the PPA expects.
- Assess before processing. Run an impact assessment for each system that processes personal data, and record the alternatives considered and the mitigations chosen.
- Write and enforce a usage policy. An acceptable use policy covers the PPA’s generative AI points, and guardrails stop personal data going where the policy says it should not.
- Report to the board. Amendment 13 and the PPA’s guidance on the board’s role make privacy oversight a board matter in organisations where data processing is central. AI belongs in that reporting.
Sector expectations on top
Israel’s sector regulators add their own expectations. An interagency taskforce of financial regulators published its final report on AI in the financial sector in December 2025, and the Israel National Digital Agency published a Guide for Responsible AI Use in the Public Sector in May 2026. Both are described on the Israel AI policy page. Israeli companies whose AI reaches the EU market may also be caught by the EU AI Act.
Where Aicura fits
Aicura supports this work. Discovery finds the AI tools in use, and the AI register records each one with the personal data it touches, its vendor and its owner. Assessments hold the impact work for each system, guardrails apply your generative AI policy to live traffic, and policies keep governance documents current. Pro adds assessment against ISO/IEC 42001 and the NIST AI RMF, which fit the risk-based approach Israel’s AI policy favours. It supports the work and the decisions stay with you.
Common questions
Is the PPA's AI guidance final?
The PPA published it as draft guidance for public consultation in 2025, with comments due by 5 June 2025. We have not found a final version as of October 2026. The PPA says it treats its guidelines as a significant regulatory tool and enforces in line with them, so the draft is the clearest statement of how it reads the law for AI.
Do I need consent to use personal data in an AI system in Israel?
In most cases the Protection of Privacy Law relies on informed consent. The PPA's draft guidance says that consent must be informed about how the AI system processes the data, among other things, and that public availability of information, such as a social media post, does not in itself amount to consent to use it for training.
Do I need a privacy protection officer if we train AI models?
Possibly. Amendment 13 requires one for public bodies, larger data brokers and organisations whose core activities involve large-scale systematic monitoring or large-scale processing of information of special sensitivity. Commentary on the draft guidance suggests that organisations training AI models on personal data are very likely to meet one of these tests.
Does Israeli law give a right to human review of AI decisions?
Not as a specific rule. The Protection of Privacy Law has no equivalent of GDPR Article 22. People have rights to access and correct their information, and the PPA expects those rights to be honoured for AI systems, including correcting inaccurate outputs.
Start with the AI you run.
Find the AI in your organisation, give every system an owner and add assessments and guardrails when the work needs them.