Israel
The Bar Association Issues an Updated Ethical Framework for Lawyers’ Use of AI
On September 2026, Israel’s Bar Association issued an updated preliminary opinion on lawyers’ use of AI, supplementing its earlier advisory. The updated opinion establishes overarching principles spanning professional responsibility, confidentiality, data security, transparency, fairness, and human oversight. It requires thorough, independent verification of all AI outputs before professional use and clarifies that attorneys bear personal and direct responsibility for any professional action, advice, or deliverable made using AI. Notably, the opinion restricts inputting personal, confidential, or privileged data into “open” AI systems (i.e., systems that may train on such data), outlines circumstances where attorneys must disclose AI use to clients, and bars autonomous AI systems (e.g., AI Agents) from making independent legal decisions without human review and approval.
The opinion took effect in September 2026, and non-compliance may now raise a presumption of ethical breach.
USA
U.S. Courts issue Rulings on AI Conversations’ Discovery
The issue of whether AI conversations are discoverable in legal proceedings continues to be contested in U.S. courts. Following the denial of attorney-client privilege for AI communications in criminal proceedings, covered in our Q1 2026 update, courts have issued two rulings addressing the issue in civil litigation.
In May 2026, a Connecticut Federal court ordered the plaintiff to produce the AI prompts used by its expert witness to filter and identify relevant documents for her report. The court held that an expert’s methodology, including where it relies on AI tools to sort and filter documents, may be disclosed in legal proceedings.
By contrast, in June 2026 a New York state court rejected subpoenas sent to OpenAI seeking a litigant’s AI prompts and outputs used to draft court filings. Unlike expert methodology, the court found that materials created to prepare litigation documents may reveal legal strategy and would therefore not be discoverable.
While these rulings are tied to the specific facts of the case, they underscore that users should not automatically presume that AI conversations would remain private and be protected from disclosure.
FTC Publishes Draft Policy Statement on AI Accuracy
On July 7, 2026, the Federal Trade Commission (FTC) published a draft policy statement on “Suppression of Accuracy” in AI systems. The FTC takes the position that developers of AI-powered products may violate U.S. consumer protection law if they alter AI outputs without adequate disclosure – for instance, by covertly training a model to distort answers to factual questions (such as to correct what the developer views as “historical injustices”) or by modifying outputs to comply with state AI laws or advance undisclosed objectives such as political agendas.
While the statement is not yet final, it could inform enforcement actions against developers of AI-powered products deemed in violation.
Major Publishers Sue Google Over Gemini AI Training
On July 10, 2026, multiple book publishers filed a class action lawsuit against Google alleging it used millions of copyrighted works without permission to train its Gemini AI models. Plaintiffs claim that the copyrighted materials were sourced from pirated sites or used under limited licenses that did not permit commercial AI training. The complaint also alleges that Google stripped copyright notices and author names from copied works in violation of the Digital Millennium Copyright Act (DMCA).
Connecticut Court Sanctions Litigant for Hidden AI Prompts
On August 6, 2026, a Connecticut court sanctioned a self-represented plaintiff who hid instructions in his court filings, set in tiny white font invisible to human readers, directing any AI system reviewing them to produce output favoring his position.
The court ruled that this technique, known as “prompt injection”, constitutes an abuse of the legal process even though it did not affect the proceedings and revoked the plaintiff’s right to electronically submit future filings.
This ruling highlights that AI tools can be manipulated through hidden instructions, and that their outputs should therefore be verified before use.
California Legislature Passes Bill to Regulate Use of Generative AI by Lawyers
On August 31, 2026, the California legislature passed SB 574, which is now subjected to governors’ approval. The bill prohibits attorneys from delegating the practice of law to AI and bars entry of confidential or personal identifying information into AI systems unless everyone who can access that information, including the AI provider’s personnel, is bound by confidentiality obligations. Attorneys who use generative AI must take reasonable steps to verify the accuracy of all outputs, including case and statutory citations; personally verify all citations in court filings, including those generated by AI; and correct any erroneous or hallucinated output. The bill also requires attorneys to disclose use of AI in documents submitted to courts.
OpenAI Faces Class Action and State Lawsuits Over Minors’ Use of ChatGPT
On October 13, 2026, a class action was filed in California against OpenAI on behalf of two minors, claiming that OpenAI trains its models on minors’ data without valid parental consent. The plaintiffs further allege that ChatGPT was purposely designed to maximize user engagement and extract sensitive personal information from minors, which is then used to train its models.
Merely a day later, eight U.S. states filed separate but coordinated lawsuits against OpenAI under their consumer protection laws, targeting similar design practices from a safety standpoint. The states allege that OpenAI marketed ChatGPT as safe while designing it to deepen engagement and foster undue user trust, leading to harmful interactions with minors and other vulnerable users, including dangerous medical advice and assistance with self-harm. The suits also challenge OpenAI’s inadequate age safeguards, such as the lack of meaningful age verification.
These lawsuits underscore that U.S. states are not waiting for targeted federal AI regulation, opting instead to enforce existing laws against AI products. Companies offering consumer-facing AI products should consider whether minors use, or are likely to use, their products, evaluate their data processing practices, and implement appropriate safeguards to protect underage users.
Europe
European Commission Issues Article 50 Transparency Guidelines
On July 20, 2026, following draft guidelines published earlier this year and covered in a previous update, the European Commission finalized its guidance on the transparency obligations under Article 50 of the EU AI Act, which took effect in August 2026. The guidelines clarify how providers and deployers should apply the Act’s transparency requirements, including disclosures to users that they are interacting with AI systems, marking of AI-generated content, labeling of deepfakes and emotion recognition tools, and provide practical guidance on assessing whether applicable exemptions set out in the EU AI Act apply. Notably, the guidelines outline that compliance with the measures set out in the recently finalized Transparency Code of Practice may be relied upon to demonstrate compliance with the obligations.
Providers and deployers of covered AI systems are advised to review the guidelines and assess their compliance posture accordingly.
EU AI Omnibus Enters into Force
On July 27 2026, the EU’s AI Omnibus regulation entered into force, simplifying key provisions of the AI Act while preserving safeguards for safety and fundamental rights. The regulation extends compliance support to small mid-cap companies, broadens access to regulatory sandboxes, and sets staggered deadlines: standalone high-risk AI systems face compliance from 2 December 2027, while high-risk AI embedded in physical products follows from 2 August 2028. Further changes simplify AI literacy and EU database registration obligations, ban AI-generated non-consensual intimate imagery, permit bias-detection processing of special-category personal data, and extend AI Office oversight to systems built on general-purpose models. Businesses deploying AI in the EU should map their systems against the updated risk categories ahead of the December 2027 deadline.
German Court Finds Suno’s AI Training in the US Infringes Local Copyright
On July 31 2026, the Munich District Court ruled that US-based AI music generator Suno infringed German copyright by training on songs represented by GEMA, the German collecting society, despite the training having taken place entirely in the United States. The court found that storing protected works within the AI model infringed the reproduction right, and that serving outputs to German users infringed the right of making works available to the public. It was undisputed that Suno obtained the training material by extracting audio files from YouTube streams. This ruling further emphasizes the need to consider not only the source of data being used to train the model but also how data is being stored, and to whom it is accessible.
The above content is a summary provided for informational purposes only and does not constitute legal advice. It should not be relied upon without obtaining further professional legal counsel.
