Israel has no dedicated AI law. An Israeli company that develops or uses AI is bound by existing law, above all the Privacy Protection Law whenever personal data is involved, and by the EU AI Act if its product is offered in the European Union or its output is used there. ISO/IEC 42001 and the NIST AI Risk Management Framework are voluntary: no law requires them, but customers, investors and partners ask about them. ISO/IEC 42001 can be certified by an external body, while the NIST framework has no certification. All three rest on the same core documents, so one well-kept set can serve as the base for all of them.
What applies in Israel
Israel's policy on AI regulation and ethics, published in December 2023 by the Ministry of Innovation, Science and Technology and the Ministry of Justice, leaves AI to the regulator of each sector, an approach it prefers to “broad horizontal legislation”. It recommends a risk-based approach and “soft” tools, such as non-binding ethical principles, standards and self-regulation. The national plan to accelerate AI (Hebrew, PDF), published by the National AI Directorate in August 2026, goes the same way: its chapter on regulation is about removing barriers, and it does not propose an AI bill.
Sector rules can still reach a supplier. The final report on AI in the financial sector (Hebrew, PDF), published in December 2025 by an inter-ministerial team that included the Ministry of Justice and the financial regulators, prefers policy principles to binding rules for now. It proposes treating a financial institution's engagement with a third party for AI systems as outsourcing, with responsibility staying with the institution and its board. Its governance toolbox is not binding until the financial regulators decide otherwise, but because the bank or insurer stays responsible, the questions can reach its suppliers.
The Privacy Protection Law, 1981 applies to every AI system that processes personal data, and the Privacy Protection Authority has published a draft guidance on how it applies. We cover it in our article on personal data in AI tools.
ISO/IEC 42001: a management system you can certify
ISO/IEC 42001:2023, published in December 2023, sets requirements for establishing, implementing, maintaining and continually improving an AI management system, for any organisation that provides or uses products or services that use AI. Like ISO/IEC 27001, it is a management-system standard: the certificate covers how the organisation manages AI, not a particular product.
The standard's table of contents shows what it asks an organisation to put in place:
- Top management commits to the system, sets an AI policy and assigns roles, responsibilities and authorities.
- The organisation assesses and treats its AI risks, and carries out impact assessments of its AI systems.
- The people involved have the competence and awareness their roles require.
- Annex A to the standard lists reference controls, which the organisation draws on when it treats its risks.
- An internal audit and a management review check that the system works, and nonconformities lead to corrective action and continual improvement.
A certificate is issued by an external certification body after an audit; ISO itself does not certify. ISO/IEC 42006:2025, published in July 2025, sets additional requirements for those bodies and can serve as the criteria for accrediting them. Accreditation is not compulsory, but when a customer asks for a certificate, it is worth checking that the certification body is accredited for this standard.
Because ISO/IEC 42001 shares the structure of other ISO management-system standards, a company that already holds ISO/IEC 27001 can build on its existing procedures, audits and management reviews. ISO/IEC 27701, the privacy standard, was reissued in October 2025 as a stand-alone standard that no longer has to sit on top of ISO/IEC 27001.
One limit matters for Europe: an ISO/IEC 42001 certificate does not by itself show compliance with the EU AI Act, which gives a presumption of conformity to harmonised standards whose references are published in the Official Journal of the European Union. On 14 October 2026 the European Commission published the reference of EN 18286:2026, on quality management systems for high-risk AI systems, and ISO/IEC 42001 is not on that list.
The NIST AI Risk Management Framework
The US National Institute of Standards and Technology published the AI Risk Management Framework 1.0 in January 2023. It is free and intended for voluntary use, and there is no certificate for it. Its core has four functions. Govern builds a culture of risk management, with policies, roles and accountability, and runs through the other three. Map sets out each system's context and the risks that arise from it. Measure analyses, tests and monitors those risks. Manage decides how to treat them and how to respond to incidents.
In July 2024 NIST added a Generative AI Profile (NIST AI 600-1). NIST states that version 1.0 is being revised as part of the White House AI Action Plan, so check for a newer version before relying on its wording.
The EU AI Act and an Israeli company
The EU AI Act, Regulation (EU) 2024/1689, reaches a company outside the EU in two main cases (Article 2(1)). The first is a provider that places an AI system or a general-purpose AI model on the EU market, or puts a system into service there, wherever it is established. The second is a provider or deployer outside the EU whose AI system produces output that is used in the EU. A provider develops a system and offers it under its own name; a deployer uses a system under its authority.
The Digital Omnibus on AI, Regulation (EU) 2026/1744, in force since 27 July 2026, moved the high-risk dates and softened some duties. The table sets out what applies from when:
| Obligation | Who it binds | Applies from |
|---|---|---|
| Prohibited AI practices | Providers and deployers | 2 February 2025. A new prohibition, on generating intimate images of an identifiable person without consent or child sexual abuse material, from 2 December 2026 |
| AI literacy of staff | Providers and deployers | 2 February 2025. Since 27 July 2026 the duty is to take measures to support it, without guaranteeing any particular level |
| General-purpose AI models | The provider of the model itself | 2 August 2025 |
| Transparency (Article 50): telling people they are dealing with an AI system, marking generated content, labelling deepfakes | Providers and deployers, each for its own duties | 2 August 2026. Generative systems placed on the market before that date must meet the marking duty by 2 December 2026 |
| High-risk systems in the areas listed in Annex III, such as recruitment, education and credit scoring | Mainly providers, and deployers for their own duties | 2 December 2027 |
| High-risk systems in products covered by the EU product laws listed in Annex I | Mainly providers, and deployers for their own duties | 2 August 2028 |
An Israeli supplier should first establish whether its product reaches the EU and in which role. A company that builds its product on another company's model is generally a downstream provider of an AI system, not the provider of the model. A chatbot or a generative feature used in the EU is already subject to the transparency duties. If a use falls within Annex III, the provider has until 2 December 2027 to put in place, among other things, risk and quality management systems and, if it is outside the EU, an authorised representative in the EU.
Choosing and combining the frameworks
Each framework answers a different demand. Where the EU AI Act applies, it is law, and no certificate replaces compliance with it. When a customer's questionnaire asks for a certificate, ISO/IEC 42001 is the AI standard that can be certified, and it extends an existing ISO/IEC 27001 system. An investor's due diligence on AI looks at the tools the company uses, the systems it develops, the risks it has identified and whether the EU AI Act applies. The NIST framework can serve as the method for the risk assessment that ISO/IEC 42001 requires, since that standard leaves the choice of risk-management method to the organisation.
Whichever framework you choose, the same documents do most of the work:
- An AI use policy says who may use which tools, for what and with what data.
- A register lists the AI tools the company uses and the AI systems it develops or sells, with an owner for each.
- A risk and impact assessment covers each system and is updated when the system or its use changes.
- A privacy impact assessment covers each system that processes personal data, as the Privacy Protection Authority's draft guidance recommends, and the GDPR requires one in certain cases.
- Defined roles say who approves a new tool or feature, and the people who use AI are trained.
- An incident procedure for AI connects to the existing security incident procedure.
This is not legal advice
This article, like everything on this site, is general information only. It is not legal advice or a legal opinion, and it does not address the circumstances of any particular organisation. The law changes, and what was accurate on the day of publication may not be accurate today, so do not rely on it without advice on your own facts. Reading it does not create a lawyer-client relationship.
The article reflects Israel's December 2023 policy on AI regulation and ethics, the national AI plan of August 2026, the December 2025 report on AI in the financial sector, the Privacy Protection Law and the Privacy Protection Authority's draft guidance on AI, the EU AI Act as amended by Regulation (EU) 2026/1744, NIST AI 100-1 and NIST AI 600-1, and the published descriptions of ISO/IEC 42001, ISO/IEC 42006 and ISO/IEC 27701. Whether the EU AI Act applies to a particular product depends on its own circumstances. Current as at 19 October 2026.