Israel - AI Regulation and Ethics

Overview

The Israeli government’s AI policy document, published in December 2023 by the Ministry of Innovation, Science and Technology in cooperation with the Ministry of Justice, defines a national approach titled "Responsible Innovation." The policy frames AI governance as a balance between stimulating innovation and upholding human rights and rule-of-law values. It prioritizes sectoral, risk-based regulation implemented by the sector-specific regulator, supported by a government-level knowledge and coordination center, and aligned where possible with leading international frameworks. The document is presented as guidance rather than binding law and explicitly favors incremental / experimental regulatory tools (e.g., sandboxes, pilot projects) and soft regulation (standards, non-binding principles), while instructing regulators to apply rigorous risk-management procedures for high-impact uses. The original government publication is available from the State of Israel publication repository and related ministry communications. For the official text (Hebrew PDF) see Policy, Regulation and Ethics Principles in the Field of Artificial Intelligence (Ministry of Innovation & Ministry of Justice, Dec 2023) and an English summary and commentary is available through ministry and sectoral press coverage including the Ministry’s synthetic descriptions and public commentary by legal and industry experts.

Definitions

The policy uses broadly applied definitions consistent with international practice. Key terms include "AI-based systems" (systems that use algorithms, models or statistical procedures to perform tasks that would otherwise require human cognitive capability), "developer" (entity designing or training models), "operator" (entity deploying or using AI systems), "user" (end-user of an AI-powered service), and "public-impact AI" (systems whose operation has material effects on individual rights, public safety, or societal functions). The document explicitly references the OECD's approach to definitions and notes that sectoral regulators may adopt more granular definitions tailored to specific fields (finance, health, infrastructure). The policy also distinguishes between lifecycle phases (design, training, deployment, monitoring) to emphasize obligations across the AI lifecycle.

Governance and Institutional Framework

The policy establishes a governance architecture built on existing ministries and sectoral regulators. Primary institutional responsibilities are assigned to: the Ministry of Innovation, Science and Technology (policy leadership and international representation), the Ministry of Justice (legal advice and compatibility with existing law), sectoral regulators (rulemaking and supervision in their domains) and proposed creation of a government "knowledge and coordination center" to consolidate competence, issue guidance, and coordinate cross-sectoral issues. The document instructs regulators to adopt risk-management frameworks and to consult the coordination center on inter-sectoral matters. The policy also recommends establishing an AI Policy Coordination Center as an inter-agency focal point to collect best practices, coordinate Israel's international engagement, support capacity building, and advise regulators on harmonization. See the official publication: Policy document (Hebrew PDF) and commentary provided by the Ministry and legal scholars summarizing the architecture.

Key Focus Areas

The policy identifies seven central challenges for private-sector AI: discrimination and bias; human oversight; explainability; disclosure of AI interactions (user notice that AI is in use); safety across lifecycle; accountability by developers/operators; and privacy/data protection. To address these, the policy recommends: (1) risk-based regulatory triage that focuses regulatory attention where impact is greatest; (2) adoption of human-centred design and human-in-the-loop measures for systems affecting rights; (3) transparency obligations in the form of meaningful user notice and documentation about capabilities and limitations; (4) technical and organizational measures to mitigate bias and discrimination, including dataset documentation; (5) safety engineering and monitoring requirements for safety-critical systems; (6) accountability documentation, logging and incident reporting mechanisms; (7) data governance, security-by-design and privacy impact assessments where personal data are processed; (8) use of regulatory sandboxes and pilot projects to allow controlled experimentation; and (9) the use of standards and voluntary codes as a first step where appropriate. The policy emphasizes aligning measures with prevailing international standards and recommends periodic reassessment to determine whether additional cross-sectoral legislation is needed.

Implementation Framework

Implementation is decentralized: sectoral regulators must translate the policy principles into concrete, domain-specific rules. The policy sets a five-part toolkit for regulators: (1) risk assessment and risk-proportionate regulation; (2) regulatory experimentation (sandboxes, pilots); (3) soft-regulatory instruments (non-binding guidance, sector codes, standards); (4) targeted binding measures for high-risk, public-impact applications; and (5) capacity building via the recommended AI Policy Coordination Center. The policy calls for practical deliverables such as regulatory guidance documents, standard operating procedures for impact assessments, technical annexes on explainability and documentation, and templates for industry registries where necessary. The policy instructs regulators to consult stakeholders (industry, academia, civil society) in drafting sectoral rules and to coordinate with the Ministry of Innovation and the proposed coordination center to promote consistency and interoperability across sectors and with international frameworks.

Monitoring and Evaluation

Monitoring is organized through a two-tier model: (A) sectoral monitoring and enforcement by sector regulators responsible for supervising compliance with domain-specific rules; and (B) an overarching monitoring and coordination role for the government knowledge center to aggregate lessons learned, produce cross-sectoral reports, identify gaps, and recommend adjustments. The policy recommends regular impact and rights assessments, reporting obligations for serious incidents affecting safety or rights, and structured evaluations of sandboxes and pilot projects to inform adaptive regulation. The document also proposes periodic review cycles to evaluate whether the sectoral approach remains sufficient or whether economy-wide/horizontal legislation should be introduced.

Penalties, Liability, and Appeals

The 2023 policy itself does not create new criminal offences or a uniform penalty regime; it is a policy guidance document that leaves enforcement to sectoral regulators and to existing legal frameworks. The policy notes that binding penalties (where appropriate) would be implemented by sectoral regulators under their statutory powers and that existing statutes (notably Israel’s Protection of Privacy Law and other sector-specific statutes) continue to apply. Subsequent legal reforms to the privacy framework (Amendment to the Protection of Privacy Law enacted in 2024) strengthen the Privacy Protection Authority’s enforcement powers and potential sanctions for privacy violations; sectoral liability (tort, contract, consumer protection) remains governed by existing law. The policy therefore focuses on compliance tools and guidance rather than prescriptive sanctions, while acknowledging that regulators may impose penalties under their statutory regimes for non-compliance.

Relationship to Other Instruments

The policy expressly situates itself alongside prior government decisions and legal instruments: it implements principles drawn from Government Decision 212 (2021) and Government Decision 173 (February 24, 2023) and is intended to operationalize elements of the National AI Program. It complements the Protection of Privacy Law (and the PPA’s guidance), existing sectoral regulatory regimes (finance, health, telecom, transport), and international commitments. The policy instructs regulators to interpret and apply existing laws to AI use (for example, privacy and consumer-protection rules) and to avoid regulatory fragmentation by coordinating through the proposed AI Policy Coordination Center. For the state publication see official PDF.

International Alignment

The policy commits Israel to international interoperability: regulators should align, when possible, with leading international frameworks (for example OECD recommendations, EU developments and other advanced jurisdictions) to reduce cross-border friction and to facilitate trade and research collaboration. The document calls for active Israeli participation in international fora, and for the national coordination center to serve as Israel’s focal point in international technical and policy cooperation. The policy’s preference for a risk-based, sectoral approach is framed as flexible to allow alignment with multiple international obligations while preserving Israel’s innovation ecosystem. The policy has been published alongside Israel’s increasing international engagement on AI safeguards and subsequent participation in global treaties and multilateral initiatives.

Implementation Timeline

Sources and References