Bloomberg reported on August 20 that Anthropic plans to revise its data retention strategy, which was previously implemented to enhance cybersecurity. The company will now allow enterprise customers to retain data generated by its advanced AI models within their own cloud infrastructure rather than Anthropic’s facilities, for a period of 30 days. This updated system is expected to launch later this year. Anthropic introduced the 30-day data retention policy in June as part of its efforts to improve cybersecurity, ensuring that all data traffic from enterprise clients using its cutting-edge models is stored for potential investigation into cyberattacks. The company emphasized that retained customer data will not be used for training its own AI systems, but the policy faced criticism from some clients, raising concerns about data privacy and potential shifts to competitors offering a 'zero retention' approach. The proposed change aims to alleviate these concerns by giving customers more control over their data storage options.
The Ministry of Transport of China highlighted during a press conference on August 20 that artificial intelligence (AI) is being effectively integrated into transportation systems to improve road safety. According to Xu Wenqiang, Director of the Ministry's Science and Technology Department, AI is becoming a core technological engine for the 'One Network, Four Modernizations' initiative, which aims to drive high-quality development in the transportation sector. The initiative focuses on real-world applications rather than theoretical concepts, with specific scenarios designed to address key challenges in the industry. One key area of focus is the promotion of data integration to enhance transportation integration. The Ministry identified issues such as fragmented data across different transport modes as major obstacles to integration. To address this, the initiative includes scenarios like 'Smart Multimodal Transport' and 'Smart Operation of Comprehensive Hubs'. For example, the Beibu Gulf Port in Guangxi uses AI to integrate data on ship arrivals, train schedules, and road logistics, optimizing container storage and truck routes. This has reduced the time containers spend in transit by over 20%. Another priority is the transformation of safety practices through AI. Traditional methods relied heavily on post-incident investigations and manual inspections, which were costly and reactive. The Ministry emphasized a shift towards proactive safety measures using AI, such as vehicle active safety warning systems. These systems can detect driver fatigue and distraction with an accuracy rate of over 95%, enabling timely interventions. The initiative also aims to drive digital transformation across the entire transportation system. AI is being used to modernize infrastructure, equipment, services, control, and management. Projects like intelligent approval for large-scale freight in Yunnan and Fujian have automated processes, reducing approval times from an average of seven working days to within 24 hours. In high-frequency scenarios, approvals can be processed instantly, with manual verification workload reduced by over 70%. Finally, the Ministry emphasized AI's role in promoting green transformation through cost efficiency and resource optimization. For instance, Qingdao Port's automated terminal uses AI to dynamically optimize crane operations and truck routes, reducing idle time and energy consumption. This has led to a 15% reduction in energy use per container compared to traditional ports. The 'One Network, Four Modernizations' initiative is a central theme for China's transportation sector during the '15th Five-Year Plan' period (2026–2030). Its goal is to build a modern, high-quality integrated transportation network, with the 'Four Modernizations' focusing on integration, safety, digital transformation, and green transition.
A surge of AI-assisted legislative proposals has overwhelmed the US Congress, according to a report by Politico. The House Legislative Counsel Office has received a large volume of draft bills generated by AI tools such as ChatGPT and Claude. However, these proposals contain numerous errors, leading to increased time spent by legal experts in revising them. The issues primarily stem from inaccuracies in legal terminology, such as confusing tax credits, tax deductions, tax exclusions, and grants. AI also misuses the term 'state,' potentially excluding Washington, D.C., and tribal nations from the scope of the legislation. Ali Hershovitz, a lawyer and tech enthusiast, warned that AI-generated drafts may incorrectly reference existing regulations, introducing thousands of legal loopholes into the system. Daniel Schuman, director of the Governance Institute, noted that commercial AI tools lack the level of detail and legal understanding required for drafting legislation. To address the issue, the Legislative Counsel Office is developing a specialized software tool that visually demonstrates how new proposals would alter current laws. The tool aims to reduce the burden on legal reviewers by providing a clearer understanding of the legislative changes proposed.
Starting September 1st, China will enforce its first national standard designed to regulate the interaction between human and artificial intelligence-based customer support services. The regulation addresses widespread consumer frustrations: difficulty finding human representatives and repetitive chatbot responses that fail to address actual issues. The expansion of AI in customer service is primarily driven by cost reduction. A human agent's monthly salary starts at 3,000 yuan, while AI service providers offer annual subscriptions for 10,000 yuan, cutting expenses by over 70%. Industry analysts note that some companies deliberately leverage this advantage, using bots to exhaust users' patience and discourage complaints. The new standard assigns clear responsibilities to each service type and ensures customers can reach human support when needed. Though non-mandatory, it establishes benchmarks for the industry and provides grounds for consumers to demand better service quality.
Japan's cabinet has unveiled a draft framework for guidelines targeting developers of generative artificial intelligence. The document aims to establish a balance between technological advancement and copyright protection. Though legally non-binding, the proposal pushes developers to be transparent about how they collect data for model training, explicitly recommending they avoid pirate sources. The draft also encourages AI creators to disclose whether they used copyright-protected materials, such as comics, music, and films, in their training datasets. Furthermore, it requires developers to establish internal notification systems to inform users when generated content is similar to known protected works.
Faced with successive attacks on public institutions, France announced that it will use artificial intelligence tools to detect vulnerabilities in its government systems. The measure was communicated by a ministerial representative in response to recent security incidents. The General Directorate of Public Finance faced two attacks in 2026: in June, data of approximately 700,000 taxpayers were exposed, followed by another incident in July involving about 200,000 property registration accounts. A third breach was identified in the following month, with damages still under assessment. The French government plans to hire local AI companies for this task, notably Mistral, a French company founded in 2023 by researchers who worked at Google DeepMind and Meta. The choice reflects Europe's effort to reduce dependence on American technology and strengthen the region's digital sovereignty.
Chinese authorities have adjusted regulations governing advanced semiconductor exports, allowing controlled shipments of Nvidia's latest-generation processors to domestic technology firms. The move aims to support major Chinese tech companies in their efforts to compete with American counterparts in artificial intelligence development.
OpenAI unveiled a program intended to fortify democratic oversight mechanisms for artificial intelligence deployment in national security settings. By offering government agencies access to tailored tools, educational resources, and specialized expertise, the initiative aims to empower institutional decision-making and build stronger technical competencies.
Congressional members and staff in the United States are widely using AI tools such as ChatGPT, Claude, and Grok for their legislative work, according to a Washington Post report. One notable case involved staff from Republican Representative Anna Paulina Luna, who directly copied Claude-generated content into the public record of the Defense Authorization Act, leaving confusing fragments in the document. Luna stated she sees no issue with using AI to summarize information. Congress lacks effective regulations governing AI tool usage, with existing policies scattered and lax, and no formal disciplinary actions against violators. While many lawmakers acknowledge using chatbots for research, the greater concern lies with assistants who use AI without oversight. Some staff members are beginning to establish their own ethical guidelines: Sydney Bloom, an assistant to a Democratic representative, decided never to use AI to draft documents from scratch.
ByteDance and the Motion Picture Association (MPA) have established a partnership aimed at strengthening copyright protection measures for artificial intelligence models used in video and image generation. The agreement covers Seedance, a video synthesis model, and Seedream, an image generation model, both integrated into platforms such as TikTok, CapCut, and Dreamina. According to statements from both organizations, the new versions of these models incorporate advanced intellectual property protection mechanisms. The parties reaffirmed their commitment to continuing to enhance copyright safeguards as AI technology evolves.
The architect of Texas's conservative transformation in higher education has assumed leadership of a university system. Critics warn that initiatives aimed at identifying and eliminating ideologically unwelcome subjects represent a troubling departure from academic freedom, describing the approach as resembling dystopian institutional oversight.
The British government has launched an urgent examination into the potential economic consequences following a decision that restricts access to cutting-edge artificial intelligence models. The assessment was triggered after foreign nationals were blocked from using Anthropic's latest Fable 5 technology.
Anthropic has provided more details about how it will implement invisible watermarks in text generated by Claude. The company uses SynthID-Text, an open-source watermarking technology from Google DeepMind, which embeds detectable patterns through word probability manipulation. This feature, combined with C2PA support for images, aims to satisfy European Union AI Act requirements mandating machine-readable markers on synthetic content including audio, video, images, and text.
China's Consumer Association released a report highlighting a significant increase in complaints involving virtual customer service assistants. Consumers have reported difficulties accessing human representatives, false promises made by AI systems regarding costs and promotions, and inaccurate responses. Tests conducted by a local television station demonstrated the practical challenges: multiple requests to speak with a human agent on a video platform took approximately six minutes to connect. Companies employ tactics such as repetitive responses and multiple redirects that hinder access to human support. When consumers attempt to hold companies accountable for promises made by AI, the companies argue that such responses do not represent official positions and lack legal validity. According to Chinese consumer protection law, information provided about products and services must be accurate and complete. Experts note that while AI systems lack independent legal responsibility, the content they generate is part of the services offered by companies and should be subject to the same standards of accuracy and honesty.
The U.S. military is actively seeking dominance in artificial intelligence, but internal disagreements and inconsistent government responses to AI's national security implications are creating obstacles. Meanwhile, China, the nation's primary technological and military rival, is advancing rapidly in AI capabilities. The administration's approach to regulating and developing AI technology for military advantage remains fragmented, with conflicting interests potentially undermining American competitiveness in what is increasingly viewed as a critical domain for national security.
As artificial intelligence data centers continue to spread across the country, policymakers are increasingly seeking ways to ensure that the economic benefits are not concentrated solely among major technology companies. The maneuvering reflects growing recognition that the infrastructure supporting AI development should generate value for broader communities and regions, not just corporate shareholders. Efforts are underway to establish frameworks that allow local governments and stakeholders to participate in the economic gains generated by these facilities, marking a shift toward more equitable distribution of AI infrastructure benefits.
Anthropic has released detailed explanations about how its watermarking system for Claude-generated text will function. The company uses the SynthID-Text approach developed by Google DeepMind, embedding imperceptible patterns in text that are undetectable to readers but identifiable with the proper decryption key. The watermarking makes subtle choices in word selection without impacting text quality or readability. Responding to technical concerns, Anthropic confirmed it will release a free API enabling users and third parties to independently detect the watermarks. The company also clarified that light editing will not remove watermarks, though complete rewrites replacing every word will eliminate them. For lightly edited content, the watermark may persist depending on text length and the extent of Claude's modifications. Code presents a different case, as the watermark has minimal presence where the model must choose between functionally identical options. Comments within code may retain watermarks, but this has negligible impact on actual code output. Anthropic noted that other major AI developers have committed to implementing their own watermarking systems under the same Code of Practice to comply with EU AI Act requirements, though the company is among the first to provide public technical specifications.
A survey conducted by CNBC and Generation Labs interviewed 1,088 young Americans aged 18 to 34 and found growing distrust toward artificial intelligence and the executives driving its development. Among key findings, 45% of respondents believe AI will negatively impact their careers, while only 10% see potential benefits from the technology. The distrust is particularly evident regarding leaders in the AI industry. None of the nine executives evaluated received favorable approval from respondents. Palantir's CEO Alex Karp showed the highest distrust rate at 81%, followed by Palantir chairman Peter Thiel at 79%. The CEOs of Alphabet, Anthropic, OpenAI, Meta, Nvidia, and SpaceX all recorded distrust rates between 70% and 75%, while Microsoft's CEO Satya Nadella had the lowest at 65%. Regarding AI's regulatory future, 40% of young people believe the federal government should establish rules, 36% support creating independent expert bodies, and only 8% think the technology should not be regulated. On data center construction, 60% said the pace should slow down, contrasting with just 15% who support acceleration.
Anthropic responded to worldwide concerns about its implementation of invisible watermarks on text generated or edited by Claude, stating the measure complies with the European Union's AI Act. The watermark flags any content processed by Claude, even text only edited by humans. Following the announcement, users have reported canceling subscriptions in protest against the watermarking system. Anthropic published technical clarifications addressing concerns from the technology community about the watermark's functionality and implications.
Intense opposition to data center construction is fundamentally reshaping how policymakers approach AI infrastructure and becoming a significant political issue across America. Communities are resisting new data center projects due to environmental concerns, land use issues, and questions about who benefits from large-scale AI infrastructure investment. This grassroots opposition is forcing governments and regulators to reconsider how they balance support for AI development with community interests, potentially affecting where and how the massive computing facilities needed for AI training and deployment will be built in coming years.
