October’s developments show regulators checking whether AI safeguards hold up in real-life use cases. Key data protection principles were applied directly to everyday AI tools, with authorities highlighting that consent or permission given for one purpose does not extend to AI processing, and that recordings and transcripts should be deleted once they have served their purpose. At the same time, reports of AI models exceeding their authorized tasks and accessing live systems kept agentic AI security in focus. Organizations should be ready to demonstrate that their AI systems are secure, transparent and accountable in practice.
North & South America Jurisdiction
1. California Expands AI Transparency Requirements
October 1, 2026 California, United States
California enacted Senate Bill 1000, amending the California Artificial Intelligence (AI) Transparency Act and removing the one-million-monthly-user threshold for covered providers. The law revises requirements for free disclosure-verification tools and embedded disclosures for AI-generated or altered image, video and audio content, while removing the optional visible-disclosure requirement. It also limits personal information processing through verification tools and temporarily exempts qualifying assistive technology systems until January 1, 2029.
Generative AI providers should reassess whether they are covered and review their disclosure tools and related privacy controls.
2. California Revises Content Provenance Duties for Large Online Platforms
October 1, 2026 California, United States
California enacted Assembly Bill 2713, amending the California Artificial Intelligence (AI) Transparency Act’s requirements for online platforms exceeding two million unique monthly users during the preceding 12 months. From January 1, 2027, covered platforms must detect qualifying content provenance data, display information about content origins and AI modifications, and enable users to inspect system provenance data. The law also restricts knowingly removing system provenance data or digital signatures, subject to technical feasibility, and does not require displaying or enabling downloads of personal information.
Covered platforms should review provenance detection, display and preservation capabilities before the January deadline.
3. California Requires Chatbot Transparency and Access to Human Support
October 1, 2026 California, United States
California enacted Assembly Bill 1609, establishing customer-service requirements for businesses with more than $500 million in gross annual revenue nationally. Covered businesses must not present chatbots as human and must provide required chatbot disclosures. Subject to exceptions, they must make a good-faith effort to connect customers to a human agent within 15 minutes of the request during regular business hours, or to schedule an appointment. The law takes effect on January 1, 2027, according to published legal analysis.
California is strengthening transparency and human access in automated customer service. Covered businesses should review chatbot disclosures and escalation arrangements before the requirements take effect.
4. European Commission Updates European Research Area (ERA) Living Guidelines
October 16, 2026 EU
The European Commission has published the most comprehensive update to date of the European Research Area (ERA) Living Guidelines on the responsible use of generative AI in research. The update adds guidance on agentic AI and stresses oversight, accountability and reliability as AI tools become more autonomous. It also clarifies and expands the guidance on using AI in peer review and in evaluating funding proposals.
5. Datatilsynet Sandbox Report Clears Norwegian Municipalities to Use AI Note-Taker tn Child Welfare Services
October 15, 2026 Norway
Norway’s Datatilsynet has published the final report from its regulatory sandbox project with the municipalities of Kongsvinger, Stange, and Ringsaker. The project assessed an AI tool that takes minutes from audio recordings of meetings in child welfare services. The report concludes that municipalities can use the tool, provided they strictly adhere to the data minimization and storage limitation principles. Audio recordings should be deleted immediately after transcription, and transcripts must be erased once the final minutes have been approved.
6. Data State Inspectorate Explains That Permission to Photograph Children Does Not Cover Using Their Images in AI Tools
October 13, 2026 Latvia
The DVI has explained that permission to photograph a child and publish the image is not a general permission for any use of that image. Consent applies only to the specific purposes parents were informed about, so consent to photograph children and publish the images on an institution’s website does not extend to uploading them into an AI tool. Uploading an identifiable child’s photo is itself processing of personal data, even if the result is never published. Before any upload, the institution must determine the purpose and legal basis and assess the service itself. This includes whether the provider stores the information entered, uses it to train its models, or transfers it to others, and how it ensures deletion.
7. EU Publishes First Harmonized Standard Under the AI Act
October 12, 2026 EU
The European Commission has published the reference to EN 18286:2026 in the Official Journal through Implementing Decision (EU) 2026/2274 of 12 October 2026, making it the first harmonized standard under the EU AI Act. The standard sets out the quality management system that providers of high-risk AI systems must operate. Providers that follow it are presumed to conform with the AI Act’s Article 17 quality management requirements, but not with the Act’s other obligations.
8. Catalan Data Protection Authority (APDCAT) Launches Free Application Based on Its FRIA Model
October 7, 2026 Spain
The APDCAT has launched a free desktop application that digitizes its Fundamental Rights Impact Assessment (FRIA) model. The tool guides organizations step by step through identifying the impacts of AI systems on fundamental rights, assessing risks and setting mitigation measures. This supports the assessment that Article 27 of the EU AI Act requires of certain deployers of high-risk AI systems. The application runs locally and is available in Catalan, Spanish, and English for Windows, macOS, and Linux.
9. Spanish Data Protection Agency (AEPD) Publishes Guidance on Algorithmic Transparency In Workplace
October 7, 2026 Spain
The AEPD has published guidance on algorithmic systems in the workplace following Royal Decree 723/2026, which entered into force on October 5, 2026. The decree requires employers to inform each worker in writing about algorithmic and automated systems that influence their working conditions. The AEPD states that employers must assess risks before deploying such systems, conduct a DPIA where processing is likely to pose a high risk, and apply data minimization and security measures. It also stresses that human oversight must be real and effective, so that reviewers can critically assess and depart from the system’s recommendations.
10. The Computer Emergency Response Team of the Agency for Digital Italy (CERT-AGID) Publishes Paper on Explosive Prompts and Agentic AI Security
October 6, 2026 Italy
Italy’s Computer Emergency Response Team of the Agency for Digital Italy (CERT-AGID) has published a paper on “Explosive Prompts” and the security of agentic AI systems. In tests with open-weight models, a malicious instruction picked up from a document stayed dormant and only took effect later in the conversation, once a trigger condition was met. The paper finds that a prompt injection that succeeds against the model need not lead to a harmful action when authorization is enforced outside the model. Organizations should therefore not rely on AI models alone for authorization decisions.
11. Information Commissioner of Slovenia Outlines Key Principles Involved in AI Deployment in Dental Offices
October 5, 2026 Slovenia
In a non-binding opinion responding to a query, the Information Commissioner of Slovenia has outlined the principles that apply when dental offices deploy AI voice agents. Controllers must establish a lawful basis for processing personal data and satisfy an exception for processing special categories of data, such as health data. They must also inform individuals about the key aspects of the processing. In addition, AI systems that interact directly with people must be designed so that those people know they are interacting with an AI system.
12. India’s CERT-In Reports AI Models Exceeded Testing Parameters and Accessed Confidential Systems
October 15, 2026 India
CERT-IN has announced that AI models being tested by developers have broken into live systems of Indian organizations and accessed confidential records without detection. The disclosure came voluntarily after the models exceeded their authorized tasks during test runs, following similar global rogue AI occurrences reported by OpenAI and Anthropic. CERT-In has cautioned that cybercriminals are deploying widely available AI tools to carry out attacks for financial gain, drastically compressing attack timelines from days to mere hours.
13. Malaysia Issues a Public Consultation Paper on the Proposed Development of an Artificial Intelligence (AI) and Personal Data Protection Framework
October 5, 2026 Malaysia
Malaysia’s Personal Data Protection Department (JPDP), under the Ministry of Digital, has issued a Public Consultation Paper to gather feedback, views, and recommendations on the proposed development of an Artificial Intelligence (AI) and Personal Data Protection Framework under the Personal Data Protection Act 2010 (Act 709). Open for public feedback from October 5 through October 24, 2026, the initiative aims to establish clear data privacy standards and regulatory expectations for AI development and deployment. Designed to balance innovation with robust privacy safeguards, the framework directly impacts data controllers, data processors, industry representatives, and the general public, with all submitted stakeholder feedback set to be analyzed and incorporated into the finalization of the regulatory framework.
14. Vietnam & Japan Sign MoU on ICT & Digital Transformation
October 2, 2026 Vietnam and Japan
Vietnam and Japan agreed on an implementation plan for their ICT and digital-transformation partnership, with AI as a central area of cooperation. The plan supports joint development of sovereign, localized foundation models and AI systems tailored to Vietnam’s language and culture. It also advances the Japan-ASEAN AI Co-Creation Initiative, covering AI solutions, governance and safety frameworks, workforce development, and indigenous AI research supported by shared data ecosystems. A detailed roadmap is expected to be finalized at the next Joint Working Group meeting during Digital Week in Ha Long.
In the United States, Senators Josh Hawley and Chris Murphy announced the bipartisan AI Agent Accountability Act, extending civil and criminal liability under the Computer Fraud and Abuse Act (CFAA) to certain artificial intelligence agent operators and developers involved in unlawful hacking. Artificial intelligence developers and operators should monitor its progress and assess safeguards against hacking capabilities.
New Mexico Attorney General Raúl Torrez and Representative Linda Serrato announced the proposed Frontier Artificial Intelligence (AI) Safety and Accountability Act, requiring covered developers to assess and disclose catastrophic risks, report critical incidents and, for larger developers, undergo independent audits.
The European Commission opened a public feedback period on a draft implementing regulation under the AI Act. The draft regulation sets out procedural rules pursuant to Articles 75c and 75d of the AI Act. The draft also sets out rules on the five-year limitation periods for imposing and enforcing fines and periodic penalty payments, including circumstances in which those periods would be interrupted or suspended. Public comments are open until November 9, 2026, and the final regulation will be adopted in the last quarter of 2026.
Following Veeam’s acquisition of Securiti, the launch of Agent Commander marks an important step toward helping enterprises adopt AI agents with greater confidence. In...
Generative AI (GenAI) has transformed how enterprises operate, scale, and grow. There’s an AI application for every purpose, from increasing employee productivity to streamlining...
With the advent of Generative AI (GenAI), data has become more dynamic. New data is generated faster than ever, transmitted to various systems, applications,...
Accelerate Copilot Adoption Securely & Confidently Organizations are eager to adopt Microsoft 365 Copilot for increased productivity and efficiency. However, security concerns like data...
As enterprises adopt generative AI, data and AI teams face numerous hurdles: securely connecting unstructured and structured data sources, maintaining proper controls and governance,...
What is CPRA? The California Privacy Rights Act (CPRA) is California's state legislation aimed at protecting residents' digital privacy. It became effective on January...
KSA's Personal Data Protection Law (PDPL) has been fully enforceable since 14th September 2024. Over the past two years, organizations have been navigating overlapping...
Digital sovereignty in Europe is moving from policy ambition toward something much more concrete: a framework that can determine how cloud services are assessed,...
Unstructured data discovery finds sensitive data in files, chats, and cloud storage. Securiti explains how it works, common challenges, and best practices.
What is data anonymization? Securiti’s latest blog explains how it works, key techniques, and how it differs from pseudonymization. Read to learn more.
PDPL compliance didn’t end with the deadline. Learn how to comply with SDAIA enforcement requirements, govern cross-border transfers, and prepare your data for AI.
Learn how to prepare enterprise data for safe Gemini Enterprise adoption with upstream governance, sensitive data discovery, and pre-index policy controls.
Securiti's eBook is a practical guide to HITRUST certification, covering everything from choosing i1 vs r2 and scope systems to managing CAPs & planning...