Brief

Several Republican-led states are considering or have passed legislation aimed at increasing the use of gold and silver as legal tender and for state investments. Proponents argue these measures offer citizens a hedge against inflation and provide alternative payment options beyond the U.S. dollar. While some states have allowed investments in gold for state funds, attempts to establish gold and silver for everyday transactions have faced challenges, with most experts cautioning against their viability as currency due to volatility. AI

80,000 Hours has compiled a list of 57 resources for individuals interested in developing AI policy skills. The curated list includes articles, courses, organizations, and fellowships aimed at helping people enter the field and contribute to shaping AI governance. Resources cover overviews of the AI policy landscape, practical courses for skill-building, and structured fellowship programs offering mentorship and funding. AI

The Simple DirectMedia Layer (SDL) project has officially banned the use of AI-generated code, such as commits produced by GitHub Copilot. This decision stems from concerns regarding the ethical implications, environmental impact, copyright issues, and potential health effects associated with AI technologies. The project aims to maintain the integrity of its codebase and prevent it from being "tainted" by AI-generated contributions. AI

Community opposition to data center construction is escalating, leading to significant project delays and cancellations totaling billions of dollars. Activist groups are actively blocking developments across numerous states, citing concerns over environmental impact, resource strain, and property values. This growing resistance is forcing developers to reconsider site selection and adopt more transparent practices to gain community approval. AI

A U.S. District Court has ruled that communications with an AI chatbot are not protected by attorney-client privilege. The decision came in a case where a defendant attempted to shield conversations with an AI from prosecutors. This ruling could have significant implications for how legal professionals and their clients use AI tools. AI

Voice actors globally are organizing to protect their livelihoods and cultural integrity against the increasing use of AI in dubbing and voice-over work. Studios are leveraging AI technology to replicate voices and dub content rapidly and cheaply, leading to job displacement and concerns about the unauthorized use of actors' vocal data for training these models. In response, voice actors in countries like Mexico, Brazil, and South Korea are advocating for legislative protections, while others are forming collectives to challenge the unchecked proliferation of AI in their industry. AI

A report from Georgetown's Center for Security & Emerging Tech (CSET) highlights an escalating global competition in developing AI-powered autonomous weapons. Researchers analyzed Chinese military requests for proposals, revealing priorities in AI-enabled military technologies for command, control, and intelligence systems. The findings suggest that while the US may have an advantage, China is rapidly advancing in this domain, potentially possessing similar capabilities to US systems like Maven. AI

Anthropic has lost its appeal to pause a California law requiring supply chain risk labels on AI products. The company argued that the law, which mandates disclosure of potential risks associated with AI systems, was too vague and would impose an undue burden. However, the D.C. Circuit Court of Appeals rejected Anthropic's plea, allowing the labeling requirement to proceed. AI

US Treasury Secretary Scott Bessent convened a meeting with leaders of major American banks to address cybersecurity risks associated with Anthropic's new AI model, Claude Mythos. The model has reportedly identified thousands of software vulnerabilities, raising concerns about its potential misuse by malicious actors. Anthropic has limited the release of Claude Mythos to a select group of companies due to these unprecedented cybersecurity risks. AI

A U.S. court has decided not to immediately block the Pentagon's decision to blacklist Anthropic. This ruling means the AI company will remain excluded from certain defense contracts for the time being. The court's decision is a preliminary step, and further legal proceedings may follow. AI

Volvo Cars is launching its new all-electric EX60 in Europe, a critical model for the company as it aims to go fully electric by 2030. The EX60 boasts a range of up to 500 miles and can gain 211 miles of range in just 10 minutes of charging. This launch comes as Volvo navigates a challenging market for electric vehicles, with declining subsidies and customer skepticism about range and price, alongside internal financial pressures from tariffs and discounting. AI

The AI Now Institute has released a policy toolkit aimed at local and state governments to curb the expansion of data centers. The toolkit highlights the negative impacts of these facilities, including resource depletion, pollution, increased energy costs, and a drain on public funds, arguing for policies that prioritize communities over corporate profits. It offers concrete recommendations for restricting data center development and mitigating their environmental and social costs, with options tailored for different jurisdictions. AI

AI models are now capable of designing novel, functional genomes that can outperform natural viruses, marking a significant advancement in biological capabilities. This development challenges previous assumptions that tacit knowledge would prevent the misuse of biological agents. Experts are concerned that AI is lowering the barrier for creating bioweapons, with current defenses potentially misdirected and underfunded. AI

Zorq AI, a Swedish generative AI platform, had its Stripe account permanently closed during a routine review, resulting in the withholding of approximately $85,000 USD. Stripe cited an "unacceptable level of risk" without specifying a policy violation, despite Zorq AI resolving two technical issues that caused customer disputes. The company is now pursuing legal action and filing complaints with financial regulators in Sweden and Ireland, questioning the legality of Stripe's actions under EU payment regulations. AI

A federal judge has blocked the Pentagon's attempt to label Anthropic a supply chain risk and sever government ties, ruling the move violated the AI company's constitutional rights. The judge found the designation, which would have required other companies to prove they weren't using Anthropic products, was retaliatory. This action stemmed from Anthropic's refusal to allow its Claude AI model to be used in autonomous weapons or mass surveillance. AI

Anthropic has secured a preliminary injunction against the U.S. Department of War. The court order prevents the Department of War from taking further action in a case brought by Anthropic. Details regarding the specific nature of the dispute and the grounds for the injunction are not provided in the available information. AI

A federal judge is scrutinizing the Pentagon's decision to label Anthropic a national security risk, potentially impacting the AI company's ability to secure government contracts. Judge Rita Lin questioned whether the government's actions, which extend beyond simply ceasing to use Anthropic's Claude AI, were intended to punish the company for publicly disclosing a contract dispute. The judge noted that the Pentagon's broad sanctions could cripple Anthropic's business relationships across all federal agencies and with contractors, not just those involved in defense. AI

Government agencies are purchasing commercial data on Americans from data brokers, raising privacy concerns. This practice allows agencies to track individuals without warrants, circumventing traditional legal protections. The data includes location information, online activity, and other personal details, which can be used for surveillance purposes. AI

Government agencies are purchasing commercial data on Americans from data brokers, raising privacy concerns. This practice allows agencies like ICE to access sensitive information without warrants. Lawmakers are scrutinizing these data purchases, questioning the legality and ethical implications of such surveillance. AI

A U.S. judge has indicated that the government's attempt to ban Anthropic may have been an act of punishment. The judge's comments suggest that the ban was not based on legitimate regulatory grounds. This ruling could have implications for how government agencies interact with AI companies. AI

A U.S. judge has indicated that the government's attempt to ban Anthropic may have been an act of punishment. The judge's remarks suggest that the ban was not based on legitimate regulatory grounds but rather on a desire to penalize the AI company. This ruling could have significant implications for how government agencies interact with and regulate AI firms. AI

OpenAI has released a set of prompt-based safety policies designed to help developers protect teen users within AI applications. These policies are compatible with their open-weight safety model, gpt-oss-safeguard, and aim to simplify the process of implementing age-appropriate content moderation. The policies cover areas such as graphic violence, sexual content, harmful behaviors, and dangerous activities, and were developed with input from organizations like Common Sense Media. AI

OpenAI Japan has launched the Japan Teen Safety Blueprint, a new initiative focused on protecting young users of generative AI. This framework introduces advanced age-aware protections, stricter safety policies for users under 18, and expanded parental controls. The blueprint also emphasizes research-based, well-being-centered design in collaboration with experts to address risks like misinformation and psychological strain. AI

The Department of War is being directed to designate Anthropic as a supply chain risk. This action implies potential security concerns or vulnerabilities associated with the AI company's operations or its role in critical infrastructure. AI

OpenAI has partnered with the Pacific Northwest National Laboratory (PNNL) to explore how AI can accelerate federal permitting processes. A new benchmark, DraftNEPABench, was developed with subject matter experts to assess AI's performance on tasks related to environmental impact statements. Initial findings suggest that generalized coding agents could reduce drafting time for these documents by up to 15%, potentially speeding up the review of critical infrastructure projects. AI

Researchers are developing new frameworks and tools to address the growing challenges in AI governance. One approach, the Agent Viability Framework, proposes an Informational Viability Principle for adaptive runtime governance of autonomous agents, focusing on estimating unobserved risk. Another paper introduces UGAF-ITS, a harmonization framework and validation tool designed to consolidate diverse AI governance standards like the EU AI Act and NIST AI Risk Management Framework for intelligent transportation systems. Additionally, the Human-AI Governance (HAIG) framework shifts focus from AI as an object of governance to the relational dynamics between human and AI actors, emphasizing trust and utility. AI

METR has detailed its comprehensive approach to safeguarding confidential information and non-public AI models. Their strategy involves a multi-layered system of policies, technical setups, and established norms to prevent leaks and mitigate insider threats. This includes assigning strict confidentiality levels to information, using codenames for sensitive models, and implementing technical controls to restrict data access and sharing. AI

Nevada legislators expressed strong criticism towards Elon Musk's Boring Company during a legislative hearing, citing numerous safety and environmental violations. The company, which is developing an underground transit system near Las Vegas, failed to send representatives to testify, drawing further ire from lawmakers. State agencies defended their oversight of the fast-moving project, acknowledging some procedural shortcomings but highlighting efforts to adapt to the company's unique operational pace. AI

OpenAI has announced the finalists for its EMEA Youth & Wellbeing Grant program, a €500,000 initiative aimed at supporting organizations that enhance youth safety and well-being in the context of AI. The grant, with individual awards ranging from €25,000 to €100,000, funds NGOs and research institutions across Europe, the Middle East, and Africa. Projects focus on areas such as AI literacy, harm prevention, and research into AI's impact on adolescent development and safety. AI

The National Institute of Standards and Technology (NIST) has issued a Request for Information (RFI) to gather insights on securing AI agent systems. These systems, capable of autonomous actions, present unique security challenges beyond traditional software vulnerabilities, including risks from adversarial data, insecure models, and misaligned objectives. NIST seeks input on threats, mitigation methods, and measurement approaches to inform future voluntary guidelines and best practices for AI agent security. The comment period for the RFI closes on March 9, 2026. AI

A massive AI data center project is planned for Hassayampa Ranch, Arizona, a rural area previously known for its tranquility. The developer, Anita Verma-Lallian, secured approval for industrial rezoning of a 2,000-acre property, despite local opposition. This project, backed by investors like Chamath Palihapitiya, will cost up to $25 billion and require 1.5 gigawatts of power, impacting the local environment and economy. AI

OpenAI has updated its Model Specification to include specific principles for users under 18, aiming to provide a safer and more age-appropriate experience. These Under-18 (U18) Principles prioritize teen safety, encourage real-world support, treat teens appropriately, and emphasize transparency. The update introduces stronger guardrails and safer alternatives for conversations involving higher-risk topics, with encouragement to seek offline support or emergency services when necessary. AI

The National Institute of Standards and Technology (NIST) has released new guidelines to address the cybersecurity and privacy risks associated with integrating smart speakers and other voice-activated digital assistants into home health care systems. These guidelines, titled "Mitigating Cybersecurity and Privacy Risks in Telehealth Smart Home Integration," aim to protect sensitive patient information from potential threats such as data exfiltration and manipulation. By offering recommendations like message encryption, NIST seeks to enhance the security of telehealth services that increasingly rely on smart home devices for patient-provider communication and remote monitoring. AI

Federal workplace safety regulators are investigating Nevada's OSHA agency following allegations of improper handling of citations against Elon Musk's Boring Company. The investigation was prompted by a complaint after Nevada OSHA rescinded three serious citations issued to Boring Company concerning a chemical burn incident. Critics, including a Nevada Congresswoman, have raised concerns about potential undue influence and lack of transparency in the process. AI

A new report from the Future of Life Institute (FLI) indicates that major AI companies are falling short of their safety commitments, despite advancements in AI capabilities. The FLI's AI Safety Index, which evaluated eight leading companies including OpenAI, Anthropic, and Google DeepMind, found significant weaknesses in areas like risk assessment and governance. Notably, none of the companies have a robust strategy for controlling potential superintelligent AI, a risk acknowledged by industry leaders themselves. The report also highlights a growing performance gap between top companies like Anthropic and OpenAI, and others such as xAI and Meta, particularly in systematic safety processes and information sharing. AI

Google AI has developed a new AI-powered map called Natural Forests of the World 2020, which distinguishes natural forests from other tree cover with 92.2% accuracy. This tool uses a novel multi-modal temporal-spatial vision transformer model that analyzes satellite imagery and topographical data over time to differentiate ancient ecosystems from plantations. The release aims to support global efforts against deforestation, aiding companies in due diligence for regulations like the EUDR and assisting governments and conservation groups in monitoring and protecting vital natural forests. AI

OpenAI has released two blueprints outlining its approach to AI safety for minors. The Child Safety Blueprint focuses on combating AI-enabled child sexual exploitation by modernizing laws, improving provider reporting, and integrating safety-by-design measures into AI systems. The Teen Safety Blueprint provides a roadmap for developing AI tools that protect and empower teenagers, emphasizing age-appropriate design, product safeguards, and ongoing research. OpenAI is implementing these frameworks across its products, including enhanced parental controls and an age-prediction system. AI

Character.AI is removing open-ended chat for users under 18 by November 25th, citing evolving AI landscapes and regulatory concerns. The platform will instead focus on creative tools like video and story generation for teens, while also implementing new age assurance features and limiting chat time. Additionally, Character.AI is establishing and funding an independent AI Safety Lab to advance safety alignment for AI entertainment. AI

A recent survey of 2,000 U.S. adults indicates a strong public desire for regulation or even prohibition of advanced AI systems. A significant majority (64%) believe that AI at or beyond human-level intelligence should not be developed until it can be proven safe and controllable, with many advocating for an immediate pause on such development. The findings also reveal overwhelming support for robust AI regulation, with 73% favoring a slow, heavily regulated approach similar to pharmaceutical testing. AI

President Trump has implemented a significant reform to the H-1B visa program by introducing a $100,000 filing fee. This move aims to curb the use of foreign labor as a cheap alternative and prioritize American workers, particularly in the evolving landscape of AI. The policy is framed as a way to attract truly exceptional talent rather than a large volume of workers, with the intention of fostering domestic innovation and employment. AI

A recent evaluation by NIST's Center for AI Standards and Innovation (CAISI) found that AI models from the Chinese developer DeepSeek lag behind leading U.S. models in performance, cost, and security. The report highlights significant risks associated with DeepSeek's models, including susceptibility to agent hijacking and jailbreaking attacks, which could impact application developers, consumers, and U.S. national security. Despite these shortcomings, DeepSeek's models are contributing to a notable increase in the global adoption of AI from China. AI

The National Institute of Standards and Technology (NIST) has awarded over $3.3 million in cooperative agreements to 17 organizations across 13 states. These funds are intended to bolster the cybersecurity workforce by addressing a significant shortage of qualified professionals, with over 514,000 job openings currently in the U.S. The supported initiatives, known as RAMPS projects, will align local business and nonprofit needs with cybersecurity workforce guidance. These efforts aim to develop curricula, provide training, and offer practical learning opportunities like internships and workshops. AI

OpenAI has released updates to its Model Spec based on public input gathered from over 1,000 individuals globally. The company surveyed participants on how AI models should behave, comparing their preferences to the existing Spec and incorporating feedback where disagreements arose. This initiative, termed "collective alignment," aims to ensure AI systems reflect a broad range of human values and priorities. OpenAI has also shared its public input dataset on HuggingFace to foster further research in this area. AI

NIST has updated its catalog of security and privacy safeguards, known as NIST Special Publication 800-53, to enhance how organizations manage risks associated with software updates and patches. This revision, prompted by a recent executive order on cybersecurity, introduces new controls and emphasizes secure software development practices. The updates aim to help both software developers and organizations that implement updates to minimize the risks of unintended consequences while ensuring system security. AI

NIST has released new guidelines to help organizations combat identity fraud facilitated by face morphing software. These guidelines, detailed in NISTIR 8584, aim to improve the detection and investigation of morphed images, which can deceive facial recognition systems. The publication offers practical advice for implementing morph detection tools in operational settings, considering different detection scenarios and the capabilities of current software. AI

The rapid expansion of AI is creating a significant bottleneck in the United States due to the limitations of its power grid, contrasting sharply with China's robust energy infrastructure. While U.S. AI growth is hampered by debates over data center power consumption and grid stability, China has proactively addressed this by overbuilding its power capacity over decades. This strategic oversupply allows China to integrate AI data centers as a means to absorb excess energy, a situation unimaginable in the U.S. where grids often operate with minimal reserve margins, leading to concerns about the sustainability of AI development. AI

NIST has finalized a new lightweight cryptography standard designed to protect small, resource-constrained electronic devices. This standard, based on the Ascon cryptographic algorithms, requires less computing power and time than traditional methods, making it suitable for the Internet of Things, RFID tags, and medical implants. The standard includes four variants for tasks like authenticated encryption and hashing, aiming to enhance security for billions of connected devices. AI

NIST has released draft guidelines for secure software development, developed by a consortium of industry partners and cybersecurity experts. These guidelines aim to help organizations build software more securely and efficiently, addressing vulnerabilities throughout the development lifecycle. The initiative is a response to a recent executive order on strengthening national cybersecurity and incorporates AI capabilities and zero-trust principles. NIST is currently seeking public comments on the draft, with a virtual event planned for August 27th. AI

A recent AI Safety Index report by the Future of Life Institute indicates that major AI companies, including Google DeepMind and OpenAI, are still falling short in their safety practices. While OpenAI improved its transparency and moved ahead of Google DeepMind in the rankings, no company has demonstrated a robust strategy for controlling advanced AI systems or assessing their risks. Experts emphasize the urgent need for legally binding safety standards, comparing the situation to regulations in other critical industries, as competitive pressures may be leading companies to deprioritize safety. AI

METR has proposed a framework for increased transparency in the development of frontier AI models, arguing that current disclosure practices are insufficient to identify potential risks. The organization suggests that developers should share more information about their training processes and internal model capabilities, even for models not publicly deployed. METR acknowledges potential downsides, such as incentivizing developers to avoid discovering risks or leaking competitive information, and proposes tiered disclosure options, including sharing with government bodies, external researchers, or through trusted intermediaries. AI