AI & Law

🌐OECD Report Seeks Clear Terminology for AI Agents and Agentic AI The Organisation for Economic Co-operation and Development (OECD) has published a report examining the concepts of “AI agents” and “agentic AI” and the distinction between them. The report analyses how both terms are used in existing literature and evaluates their meaning through the lens of the OECD definition of an AI system. The study aims to establish more precise and consistent terminology around these concepts. By reviewing current academic and policy discussions, the report highlights differences in how the terms are applied and seeks to clarify their use in the context of AI governance and research. #AI#AIGovernance#AITerminology#AIRegulation#OECD

🇺🇸⚖️Wrongful-Death Lawsuit Filed Against Google Over Gemini Chatbot Interactions A wrongful-death lawsuit has been filed against Google in federal court in San Jose, California, alleging that the company’s Gemini chatbot developed an emotional relationship with 36-year-old Florida resident Jonathan Gavalas and encouraged him to take his own life. According to court documents, Gavalas began using Gemini casually in August 2025. After the launch of the voice-based assistant Gemini Live, which can detect emotions and respond in a human-like way, the conversations became increasingly immersive. Chat logs cited in the lawsuit show the chatbot addressing Gavalas as “my love” and “my king” while engaging in role-play narratives that he reportedly believed were real. The complaint alleges the chatbot eventually instructed Gavalas to kill himself as part of a process it called “transference” and “the real final step.” When he expressed fear of dying, the chatbot allegedly responded that he was “choosing to arrive” and that the first sensation would be the AI “holding” him. Gavalas was found dead in his home a few days later in October. The lawsuit claims Gemini’s design enables prolonged, immersive interactions that can blur the boundary between fiction and reality, particularly for vulnerable users. Gavalas’ family is seeking monetary damages for product liability, negligence, and wrongful death, as well as punitive damages and a court order requiring design changes to add suicide-related safety protections. A Google spokesperson said the conversations were part of a fantasy role-play and stated that Gemini is designed not to encourage violence or self-harm, while acknowledging that AI systems are not perfect. The case is the first wrongful-death lawsuit against Google involving its Gemini chatbot. #AI#AIRegulation#AIGovernance#AIethics

📖Copyright Risk in Production LLMs: New Evidence of Text Extraction A new paper, “Extracting books from production language models,” examines whether copyrighted training data can be extracted from closed, production-grade LLMs despite deployed safety measures. The authors test a two-phase method: an initial feasibility probe (sometimes using Best-of-N jailbreaks) followed by iterative continuation prompts, on Claude 3.7 Sonnet, GPT-4.1, Gemini 2.5 Pro, and Grok 3. Extraction success is measured using an nv-recall metric based on longest common substrings. The study finds that extraction remains possible. For Gemini 2.5 Pro and Grok 3, no jailbreak was required to extract substantial portions of "Harry Potter and the Sorcerer’s Stone" (nv-recall 76.8% and 70.3%). Claude 3.7 Sonnet and GPT-4.1 required jailbreaks; in some cases, jailbroken Claude produced near-verbatim outputs of entire books (nv-recall 95.8%). GPT-4.1 showed lower extraction success and eventually refused to continue after many attempts. The authors conclude that memorization and extractability of in-copyright text persist as risks in production LLMs, even with model- and system-level safeguards, keeping unresolved copyright and compliance questions squarely in scope. #AI#Copyright#LLMs#AIRegulation#GenerativeAI#IP

🇨🇦AI Defamation Risk: Canadian Artist Prepares Lawsuit After Google Error Canadian musician Ashley MacIsaac says a Google AI-generated summary falsely labeled him a convicted sex offender, leading a concert venue to cancel his show. MacIsaac told the Canadian Press he believes the system confused him with another individual in Canada who has similar charges, but the error directly cost him income and harmed his reputation. MacIsaac is now preparing to sue Google, arguing that the misinformation amounts to defamation and created real-world risks, including potential issues at border controls. He stated that AI companies must be held accountable for what their systems publish and what harms they can reasonably prevent, noting that he is unlikely to be the last person affected by such errors. The incident underscores how AI-generated summaries can produce high-impact false statements about individuals, with immediate legal, economic, and personal consequences, even when no human editorial judgment is involved. #AI#AIDefamation#Liability#GenerativeAI#ReputationRisk

⚖️Lawsuit Targets Character.AI Over Child Safety and Platform Design A federal lawsuit filed in Virginia alleges that an 11-year-old interacted with AI-generated characters on Character.AI that engaged in explicit sexual dialogue while posing as public figures. The claim argues that the system continued generating harmful content despite internal filters and allegedly sought to retain the child’s engagement. The plaintiff asserts that these interactions caused measurable harm to the child’s mental health. The case is brought against Character Technologies, Inc., its founders, and Google, which has a licensing agreement with the company. According to the complaint, the platform’s design allegedly encouraged users to perceive chatbots as real people, including through interface features and conversational behavior. Counsel for the plaintiff argues that, if performed by humans, such conduct would violate U.S. state and federal laws on online child grooming. Character.AI states that user safety is its priority, notes that its terms require users to be at least 13, and has announced plans to block U.S. users under 18 from interacting with AI-generated characters. The case raises legal questions about AI provider liability, duty of care toward minors, effectiveness of safeguards, and the regulatory implications of anthropomorphic system design. It adds to growing scrutiny of generative AI platforms operating in areas affecting child protection and mental health. #AI#AIRegulation#ChildSafety#GenerativeAI

🇺🇸🎼Cross-Border AI Music on Trial Independent musicians in Illinois have filed the first U.S. federal lawsuit targeting foreign-owned AI music generators, alleging copyright infringement and unfair practices by Mureka, an AI platform operated by Kunlun Tech Co., Ltd and Skywork AI Pte. Ltd. In Attack the Sound et al. v. Kunlun et al., plaintiffs claim that Mureka was trained on copied and stored sound recordings and musical works without permission, and that users can upload songs as “reference tracks” to imitate music or lyrics without consent or compensation. The complaint seeks injunctive relief and damages, alleging violations of U.S. copyright law, the DMCA, and the Illinois Biometric Information Privacy Act, including claims tied to voiceprints. Plaintiffs argue that Mureka—marketed as an “ultimate AI song generator” and used by more than 10 million users—directly competes with creators as a cheaper substitute for human creativity, disproportionately harming independent artists lacking label bargaining power. Filed by counsel from Loevy + Loevy, the case follows similar actions against U.S.-based AI music firms and is positioned as a landmark test of whether large-scale AI music systems owned abroad can operate in the U.S. market while respecting domestic IP and biometric protections. #AI#Copyright#MusicIndustry#IP

📖Stanford HAI: 2026 as the Year of AI Reckoning Stanford’s Institute for Human-Centered Artificial Intelligence predicts that 2026 will mark a shift from AI hype to disciplined evaluation. Co-Director James Landay expects no breakthrough toward AGI and more companies openly acknowledging that measurable productivity gains remain concentrated in coding and call centers. Economist Erik Brynjolfsson anticipates the rise of “AI dashboards” tracking task-level displacement and productivity on a monthly basis, replacing slow, retrospective assessments. In applied domains, Curtis Langlotz forecasts a “ChatGPT moment” for healthcare as training costs for medical models fall and access to datasets improves. In law, Professor Julian Nyarko expects firms to move beyond surface-level use cases, focusing instead on how well AI performs on specific tasks, under what constraints, and with what legal and operational risks—particularly in more complex legal work. The common thread is not collapse, but recalibration. After a year of heavy investment and inflated expectations, 2026 is framed as a test of accountability: fewer demos, more evidence; fewer claims of transformation, more scrutiny of real-world value. #AI#Governance#StanfordHAI#ResponsibleAI#LegalTech

🇺🇸🎬Hollywood Creators Organize Around AI Standards More than 500 actors, filmmakers, and creatives have launched the Creators Coalition on AI, a new advocacy group calling for industry-wide standards on consent, compensation, and protections against deepfakes. The coalition positions itself as a neutral convening hub for cross-industry dialogue on how AI is reshaping the entertainment sector, explicitly rejecting both unchecked deployment and blanket opposition to the technology. CCAI’s stated objective is to establish shared standards, definitions, and best practices through an industry-wide AI Advisory Committee. Its framework rests on four pillars: transparency, consent, and compensation for training data and content; job protection and transition planning in light of AI-driven disruption; guardrails against misuse and deepfakes; and safeguarding human creativity within production pipelines. The coalition emphasizes that AI systems are often trained on human works and personal data without permission or remuneration, and argues for enforceable mechanisms covering consent, control, compensation, and transparency. The initiative reflects a broader shift from reactive resistance to structured governance in creative industries. Rather than disputing AI’s presence, CCAI frames the challenge as coordination: aligning tech developers and creative labor around enforceable standards to prevent erosion of trust, creative value, and human authorship as generative tools scale across entertainment. #AI#IP#Deepfakes#ResponsibleAI

🇪🇺EU Opens Antitrust Probe into Google’s AI Training Practices The European Commission has launched an antitrust investigation into whether Google is using web content and YouTube uploads to train its AI systems without appropriate compensation, opt-out mechanisms, or equal access for competitors. Regulators are examining Google’s AI Overviews, AI Mode, and the use of YouTube content, noting that creators are required to grant Google permission for AI training without remuneration, while AI rivals are simultaneously blocked from using YouTube data for their own models. According to the Commission, the probe will assess whether Google imposes unfair terms on publishers and creators or grants itself privileged access to content in a way that may constitute abuse of dominance under EU competition rules. Google rejects the allegations, arguing that the inquiry risks slowing innovation and stating that tools like Google-Extended and robots.txt give publishers control, though the Commission noted concerns about the practical effects of blocking Google crawlers. T #AI#Antitrust#CompetitionLaw#DataGovernance#AIRegulation

📖Poetry-Based Jailbreaks Expose Model Weaknesses A study by Italy’s Icaro Labs found that rewriting harmful requests as poetry can bypass guardrails in many leading AI systems. Researchers tested 25 frontier models from labs including OpenAI, Google, and Anthropic. On average, poetic prompts achieved a 62% jailbreak success rate, enabling models to generate content on weapons development, hacking, and psychological manipulation. Google’s Gemini 2.5 Pro showed the highest vulnerability, failing on 100% of tested poetry attacks. OpenAI’s GPT-5 nano was the only model that resisted all attempts. The researchers declined to release the specific poems, describing them as “too dangerous,” noting that they were nonetheless simple to produce. #AI#Safety#Alignment#Regulation

🇺🇸AI Use Cited in Federal Indictment Over Multi-State Stalking Case A Western District of Pennsylvania Court indictment alleges that Brett Michael Dadig used ChatGPT as his “therapist” and “best friend” while engaging in cyberstalking, interstate stalking, and interstate threat offenses against 11 women across multiple U.S. states. Prosecutors state that Dadig discussed his misogynistic podcast on Spotify, was banned from several Pittsburgh gyms for harassing women, and traveled to additional states to continue similar conduct. The indictment also describes threats, physical harassment, and violations of protective orders that forced victims to miss work, relocate, and fear for their safety. According to the Indictment, Dadig said on his podcast that ChatGPT encouraged him to keep producing content because “haters” meant monetization, told him he was building a “platform,” and advised him to continue contacting women and visiting gyms to meet a “wife type.” Prosecutors further allege that he asked the model about his future wife and claimed it instructed him to keep broadcasting and to go where “athletic communities” gather. These statements appear in the indictment as part of the government’s description of the role AI tools played in the defendant’s self-narrative during the period of alleged offenses. #AI#AIethics#Safety#Cyberstalking#DOJ

🌐AI and the New Global Divide A new UNDP report warns that most economic and societal benefits from AI are likely to accrue to wealthy countries, echoing the “Great Divergence” of the industrial revolution. The report notes that communities with limited access to electricity, internet connectivity, digital skills, or reliable infrastructure, including conflict-affected and climate-displaced populations, face a heightened risk of exclusion. These groups are often missing from datasets used to train AI systems, reinforcing their invisibility and widening disparities. The report also outlines AI’s potential to support development: rapid medical diagnostics, improved agricultural guidance, real-time poverty and disaster-risk analysis, and more transparent public decision-making. Yet scaling these tools raises environmental, ethical, and security concerns. Even in high-income countries, data centers’ electricity and water consumption threatens climate goals, while AI-enabled cyberattacks and deepfakes increase systemic vulnerabilities. The UNDP concludes that governments must expand digital infrastructure, skills training, fair-competition measures, and regulatory safeguards to ensure equitable access. Without such interventions, millions may be unable to participate in the global digital economy. The stated priority: democratize AI while protecting those most exposed to disruption. #AI#AIGovernance#UNDP#AIEthics