Намери подобно съдържание

Regulations Unveiled for AI Watermarking Hello eeverybody! Tambiama André Madiega, Policy Analyst at the European Parliamentary Research Service, illuminates the evolving landscape of generative AI in a comprehensive briefing. Madiega paints a vivid picture of the concerns surrounding generative AI, including plagiarism, privacy invasion, AI hallucination leading to the dissemination of false information, copyright infringements, the specter of disinformation and the challenge of distinguishing AI-generated content from its human counterparts. #AILaw#AIAct

Understanding Concerns about EU AI Act Hello, everybody! In a recent blog post, German doctoral students Anton Leicht and Dominik Hermle dissected the ongoing criticism surrounding the EU AI Act's regulations on foundation models. The focus of the debate has been the potential economic impact, particularly on the development of European foundation models. Leicht and Hermle underscore that key EU players, like Aleph Alpha and MistralAI, have garnered significant investments, yet lag behind global counterparts such as GPT-3.5 and GPT-4 in both performance and applications. Despite securing hundreds of millions in funding, Aleph Alpha and MistralAI's top models still fall short of GPT-3 and Meta's LlaMa-2 weakest version. The crux of the issue lies in concerns about economic competitiveness against giants like OpenAI, Google, and Meta. Critics, notably from France and Germany, fear that stringent regulations might impede European foundation model development, hindering the continent's ability to catch up with global AI leaders. Leicht and Hermle caution against dismissing comprehensive regulations. They argue that avoiding stringent oversight may risk burdening downstream deployers, potentially leading to increased compliance costs and economic pitfalls. The debate on finding the right balance between regulation and fostering innovation in the AI landscape continues. #AILaw#EUAIAct

🇺🇸Senate’s AI Jobs Report Built on ChatGPT Data Raises Questions About Policy Integrity A new report from the U.S. Senate Committee on Health, Education, Labor, and Pensions (HELP), led by Senator Bernie Sanders, claims that AI and automation could replace 97 million American jobs over the next decade. However, the report’s startling numbers, such as 89% of fast-food workers and 83% of customer service representatives at risk, were generated not by economists or federal research, but by ChatGPT. Senator Sanders used these figures to call for sweeping reforms, including a “robot tax,” a 32-hour work week without pay cuts, and a ban on stock buybacks. Yet experts note that much of the underlying data appears unreliable, with no current evidence that AI is displacing workers at this scale. Recent studies, including one from Yale, found no discernible disruption in the labor market attributable to AI. #AIPolicy#AIlaw

США: Бюро по авторским правам отклонило запрос художника Привет, сообщество AI & Law! Работа Мэтью Аллена "Пространственный театр с искусственным интеллектом" (Théâtre D'opéra Spatial) заняла первое место на артистической ярмарке штата Колорадо. Однако Бюро по авторскому праву США вынесло вердикт: авторское право на нее не распространяется, поскольку она является скорее "машинной", чем "человеческой". Для создания своего шедевра Аллен использовал генеративную программу ИИ Midjourney. Однако защита авторских прав не распространяется на произведения, созданные с помощью ИИ даже если результат производит впечатление на знатоков искусства. Это решение согласуется с предыдущими постановлениями, признающими только авторствр человека. Эта история напоминает 2018 год, когда фотография, сделанная обезьяной не получила защиту авторских прав, поскольку обезьяны не люди. Получается, что сейчас ИИ и животные имеют аналогичный авторский статус. Аллен упорно пытался зарегистрировать свою работу, подчеркивая, как сильно он настраивал Midjourney, корректировал необработанные изображения с помощью Adobe Photoshop, а также предоставлял сотни текстовых подсказок и правок. Хотя Бюро по авторским правам признало его усилия по изменению отдельных частей произведения, оно отказало в защите авторских прав на произведение в целом. В окончательной апелляции, поданной в июле этого года, Аллен утверждал, что Бюро по авторским правам не учло "существенный элемент человеческого творчества", необходимый для Midjourney. Он ссылался на доктрину добросовестного использования, утверждая, что его работа представляет собой трансформационное использование материалов, защищенных авторским правом. Однако Бюро сохранило свою позицию: произведение не может быть зарегистрировано. Теперь Аллен готовится оспорить это решение в суде, намереваясь установить права на искусство, созданное с помощью ИИ. Следите вместе с нами за развитием событий на канале ИИ & Право. #AILaw#Copyright#AIArt

AI Transparency in Focus: The iBorderCtrl Decision Hello, everybody! On September 7th, 2023, the Court of Justice of the European Union (CJEU) made a significant decision regarding the iBorderCtrl project. This initiative, funded by the European Commission since 2016, aims to use AI-powered "lie detectors" to automate security processes at EU external borders. By asking travelers a series of questions and analyzing their facial expressions, the system determines potential threats. Member of the European Parliament Patrick Breyer sought access to documents related to the iBorderCtrl project, raising concerns about its ethics and legality. However, full access was denied, citing potential harm to the consortium's commercial interests, including intellectual property rights. The CJEU upheld the decision that limited public access to project documents. It recognized the public interest but deemed that disseminating the results was sufficient. This case raises questions about transparency in AI system development, especially during the research phase. The EU's proposed AI Act acknowledges the need for transparency but excludes AI systems used for biometric categorization related to criminal offenses from this obligation. The verdict highlights the ongoing tension between transparency, public interest, and commercial interests in AI projects. #AILaw#Transparency#EUAIAct

📖AI Openness: OECD Outlines Benefits, Risks, and Policy Trade-offs The OECD has released “AI Openness: A Primer for Policymakers”, providing a structured analysis of AI openness and its implications for governance. The report argues that the term “open source” fails to fully capture the complexity of AI systems, where openness spans model weights, code, data, and documentation. It introduces the concept of open-weight foundation models, AI systems with publicly available trained weights, now comprising over half of the market, with the U.S., China, and France leading development. The report emphasizes both benefits and risks: open-weight models can accelerate innovation, competition, and accountability but also raise concerns over deepfakes, cyberattacks, and misuse in sensitive domains. Policymakers are urged to weigh marginal benefits against emerging risks within a broader risk governance framework, particularly as falling compute costs make both legitimate and malicious use more accessible. #AILaw#AIEthics#OECD#AIGovernance

Washington Judge Blocks AI-Enhanced Video as Evidence A Washington state judge presiding over a triple murder case has prohibited the use of video enhanced by artificial intelligence as evidence. This ruling, signed by King County Superior Court Judge Leroy McCullogh, marks a potential precedent in US criminal courts. The judge described AI-enhanced video as novel and cited concerns over its opaque methods and potential to confuse eyewitness testimony. The case involved lawyers seeking to introduce cellphone footage enhanced by machine learning software. Prosecutors argued against its admissibility, noting the lack of legal precedent and the enhanced video's inaccuracies. The defendant, Joshua Puloka, claims self-defense in the killings, captured on the cellphone video. The technology used for enhancement was deemed unsuitable for forensic or legal applications. Legal experts and forensic analysts highlighted the lack of peer-reviewed methodologies for AI video enhancements, emphasizing the need for further research before its use in legal proceedings. #AIlaw#ML

India's TRAI Reveals AI Regulation Recommendations Hello, everyone! India's Telecom Regulatory Authority (TRAI) has released groundbreaking recommendations on "Leveraging Artificial Intelligence and Big Data in Telecommunication Sector" following a thorough consultation process. The recommendations highlight the "urgent need" for a comprehensive regulatory framework covering all sectors. They emphasize regulating high-risk AI use cases through legally binding obligations. TRAI proposes the creation of an independent statutory authority to oversee responsible AI development and usage. This authority would have the power to establish regulations, design an AI governance framework, formulate ethical codes, and oversee enforcement mechanisms. #AILaw#AI#India

🇻🇳Vietnam Discusses Draft AI Law Inspired by EU Model Vietnam is accelerating its move toward comprehensive AI regulation. In late September, the Ministry of Science and Technology unveiled the country’s first dedicated AI law, drafted just weeks after receiving the government’s mandate. The legislation aims to create a unified legal regime that consolidates oversight, replaces fragmented rules, and aligns AI governance with both national priorities and global standards. The draft mirrors the EU AI Act with a risk-based classification of AI systems into four levels, from “unacceptable” to “low-risk,” and imposes strict obligations for high-risk applications in sectors such as health, finance, education, and justice. It also introduces regulation for general-purpose AI models, role-based accountability across the AI supply chain, and revenue-based penalties for violations. Unlike the EU, Vietnam pairs regulation with direct innovation incentives, including a National AI Development Fund, regulatory sandboxes, and tax breaks for startups and high-tech clusters. The proposed law will be rolled out in phases starting January 2026, with full implementation of high-risk AI obligations by mid-2027. With public consultation closing on 20 October, this fast-track timeline signals Vietnam’s ambition to position itself as a leader in AI governance while preserving national sovereignty and integrating global best practices. #AILaw#AIRegulation#Vietnam

🇳🇱Netherlands Issues Practical Guidance on EU AI Act Compliance The Dutch Ministry of Economic Affairs has released a detailed report outlining how organizations can comply with the EU AI Act. The document clarifies requirements for general-purpose AI models, high-risk AI systems, and the safeguards needed to ensure data protection and transparency. By breaking down obligations across risk categories, the guidance aims to help companies navigate the EU AI Act’s technical and legal standards more effectively, ensuring that deployment of AI systems aligns with both regulatory and ethical expectations. #EUAIAct#AIlaw#EUregulation

Former Trump Lawyer Cites Fake AI-Generated Cases Hello, everyone! Michael Cohen, former lawyer to Donald Trump, inadvertently included fabricated court cases in a legal document presented before a federal judge. The cases were generated by Google's AI chatbot, Bard, which Cohen apparently confused for a highly advanced search engine. Cohen sought to reduce his three-year probation, but the judge revealed that "none of these cases exist." In response, Cohen explained that he used Google Bard for legal research, assuming it was a robust search engine. Unaware of the potential for AI-generated content, he sent his findings to his lawyer, David Schwartz, who included the fictitious citations in the motion. Now, Schwartz faces potential sanctions for the inclusion of these false references. #AILaw#LegalTech#AIandLaw#LegalAI

Copyright Office Rejects Artist's Claim Greetings AI & Law community! Matthew Allen's AI artwork, Théâtre D’opéra Spatial, claimed first prize at the Colorado State Fair. However, the US Copiright Office has delivered a verdict: it can't be copyrighted due to being more "machine" than "human." Allen employed the generative AI program Midjourney to craft his masterpiece. Yet, copyright protections do not extend to artificial intelligence, even when it impresses art connoisseurs. This decision aligns with previous rulings that demand human authorship. This story harkens back to 2018 when a photograph taken by a macaque monkey was declared public domain as monkeys can't claim copyrights. Similarly, under current law, machines and animals share a similar copyright status. Allen persistently tried to register his work, emphasizing how much he had tweaked Midjourney's output, adjusted raw images with Adobe Photoshop, and provided hundreds of text prompts and revisions. While the Copyright Office acknowledged his efforts in altering parts of the artwork, it denied copyright protection for the entire piece. In a final appeal this July, Allen contended that the Copyright Office disregarded the "essential element of human creativity" required for Midjourney. He invoked the fair use doctrine, asserting that his work constitutes a transformative use of copyrighted material. However, the Copyright Office upheld its stance: the artwork cannot be registered. Allen is now gearing up to challenge this ruling in court, determined to establish the rights of AI-generated art. #AILaw#Copyright#AIArt