Regulatory briefing
On 8 October 2026, Anthropic announced a new version of its usage policy, applicable since 12 November 2026. The text is binding on organisations that use its models, directly or through the products they build. It introduces specific requirements for recommendations made in high-risk domains and for chatbots intended for the public.

The text of the usage policy
The usage policy has three parts. First, it sets universal usage standards, which list prohibited uses. These include violating the law, compromising critical infrastructure or computer systems, invading privacy, disinformation and fraudulent practices, among others. Second, it sets requirements specific to high-risk uses. Finally, it adds guidelines for certain use cases. These cover consumer-facing chatbots and AI agents, products intended for minors, agentic uses and software listed in the company’s directories.
High-risk recommendations under the usage policy
The policy covers eleven domains. They comprise the legal, medical, finance, credit, insurance, housing and employment domains. They also comprise education and qualifications, access to healthcare, public benefits and services, legal status and judicial decisions. In these domains, a recommendation produced by AI is subject to two requirements:
- effective review by a qualified person, authorised to modify it, before its delivery or use;
- clear information on the use of AI to produce it. The recipients are the persons receiving the advice or subject to the decision. There is no obligation to name the company or the model.
Recommendations entirely favourable to the person may, where the law permits, proceed without prior human review. This exception covers four listed situations. These are payment of an insurance claim and approval of coverage or prior authorisation for care prescribed by a clinician. They also include grant of eligibility for a health service or programme. Finally, they include grant or continuation of a public benefit or service. However, the exception excludes partial, reduced or conditional approvals, and the disclosure obligation may remain.
In addition, specific requirements concern physical actions. A qualified person must be able to observe the equipment and stop it. Moreover, mechanisms independent of the model’s outputs must ensure the operating limits.
Chatbots and agentic uses
Operators of consumer-facing chatbots and AI agents must tell users that they are interacting with an AI, not a human. This information appears, at a minimum, at the start of each session or in the product interface. In addition, for agentic uses, the user is responsible for the compliance of the agents it builds or deploys. That responsibility includes actions carried out by means of tools, browsers or connected systems.
Implications: the usage policy and applicable law
These requirements converge with the applicable law. Thus, information on the nature of the interlocutor matches Article 50(1) of the EU AI Act. Similarly, human review of decisions matches Article 11 of Law No. 09-08. That article restricts decisions based solely on automated processing. A Moroccan organisation may build a customer service agent, a recruitment support tool or an advisory assistant on these models. In that case, it must incorporate from the design stage the information of users, qualified review and traceability of actions.
Other providers have their own usage policies, whose content differs. Therefore, the applicable usage policy requires checking for each provider and at each amendment.
Points of attention
- Identify uses falling within the eleven high-risk domains.
- Provide for review by a qualified person before delivery of any recommendation in these domains.
- Inform the persons concerned of the use of AI in the advice and decisions concerning them.
- Display the AI nature of any chatbot from the start of the session.
- Frame the permissions and record the actions of deployed agents.
References
- Anthropic, Usage Policy, version applicable as at 12 November 2026, https://www.anthropic.com/legal/aup (page consulted on 21 October 2026).
- Anthropic, announcement of 8 October 2026 on the update to the usage policy, https://anthropic.com/news/2026-usage-policy-update (page consulted on 21 October 2026).
- Regulation (EU) 2024/1689 of the European Parliament and of the Council of 13 June 2024 laying down harmonised rules on artificial intelligence and amending Regulations (EC) No 300/2008, (EU) No 167/2013, (EU) No 168/2013, (EU) 2018/858, (EU) 2018/1139 and (EU) 2019/2144 and Directives 2014/90/EU, (EU) 2016/797 and (EU) 2020/1828 (Artificial Intelligence Act), OJ L, 2024/1689, 12.7.2024, Article 50(1).
- Law No. 09-08 on the protection of individuals with regard to the processing of personal data, promulgated by Dahir No. 1-09-15 of 22 Safar 1430 (18 February 2009), Official Bulletin No. 5714 of 5 March 2009 (official translation edition), Article 11.
See also: “Agentic AI” offered to MSMEs: what the offers cover; Guide – Configuring an AI assistant; Checklist – Compliance of AI uses. Information Systems & Digital Transformation practice.
General information document; it constitutes neither legal advice nor an opinion on a particular situation. References to legal texts should be checked against the version in force on the date of use.
Consulting firm not registered with any Bar; no representation before the courts and no act reserved to the regulated legal professions.
