Should AI Education Be Mandatory for Every U.S. Student? Over 250 CEOs Think So!
LinkedIn reveals AI engineering as the fastest-growing job for graduates, while 250 CEOs advocate for mandatory AI education in U.S. K-12 schools to bridge the skills gap...
Today's highlights:
You are reading the 93rd edition of the The Responsible AI Digest by SoRAI (School of Responsible AI) . Subscribe today for regular updates!
At the School of Responsible AI (SoRAI), we empower individuals and organizations to become AI literate through comprehensive, practical, and engaging programs. For individuals, we offer specialized training like AI Governance certifications (AIGP, RAI) and an immersive AI Literacy specialization, including our flagship 20-hour live online course starting May 25th, 2025. This structured training spans five domains—from foundational understanding to ethical governance—and features interactive sessions, a capstone project, and real-world AI projects. For organizations, including educational institutions, workplaces, and government agencies, our tailored programs assess existing AI literacy levels, craft customized training aligned with organizational needs and compliance standards, and provide ongoing support to ensure responsible AI deployment. Want to join? Apply today as an individual: Link; Apply as an organization: Link
🚀 AI Breakthroughs
LinkedIn Provides Insight into Top Opportunities for Graduates
• LinkedIn's data reveals that AI engineering is the fastest-growing job for recent graduates, with an emphasis on exploring diverse career opportunities for job seekers facing a tough market;
• Recent grads are increasingly drawn to hands-on roles in emerging industries like construction, utilities, and mining, fueled by tech adoption and infrastructure investments;
• Fast-growing cities like Tucson and Savannah are becoming prime destinations for new grads, offering affordable living and vibrant opportunities beyond traditional tech hubs.
Over 250 CEOs Urge Inclusion of AI in K-12 Education Curriculum Nationwide
• More than 250 CEOs, including top tech executives, called for integrating AI and computer science as core K-12 subjects, claiming it's essential for U.S. competitiveness
• The open letter highlighted AI and computer science education's potential to close wage gaps and unlock $660 billion annually for Americans by addressing current skills shortages
• Despite unanimous support, only 12 states mandate basic computer science education, prompting concerns that the U.S. lags behind nations like Brazil, China, and South Korea.
OpenAI Transitions to Public Benefit Corporation to Enhance AGI Accessibility and Impact
• OpenAI plans to transition its for-profit entity to a Public Benefit Corporation (PBC), ensuring alignment with its mission to benefit humanity and considering both shareholder interests and social objectives;
• While the nonprofit will maintain control and become a major shareholder in the PBC, it aims to use increased resources to enhance AI's positive impact across various communities worldwide;
• OpenAI underscores its commitment to democratic AI by advocating for AGI tools that empower humanity, balancing freedom with ethical usage, and striving to make AI accessible for all.
Netflix Enhances TV Interface with Modern Design and AI-Powered Search Features
• Netflix unveils a revamped TV interface featuring a more intuitive design and real-time recommendations tailored to user moods and interests, enhancing the overall viewing experience
• A new feature allows users to make quick, informed viewing decisions with prominent highlights of notable achievements like "Emmy Award Winner" displayed directly on the selection screen
• Alongside visual upgrades, a beta testing phase for a generative AI-driven search on iOS introduces natural language queries, offering a conversational way for users to discover content.
Hugging Face Launches Open Computer Agent to Operate Virtual Linux Machine Remotely
• Hugging Face unveiled Open Computer Agent, a free, cloud-based AI tool allowing users to operate a virtual Linux computer via text prompts in any browser
• Underpinned by smolagents, Qwen2-VL-72B, and E2B Desktop, the agent performs standard browser interactions like visiting websites or finding directions but struggles with complex tasks
• Featuring common applications like Firefox, the tool has encountered slower response times and CAPTCHA difficulties in tests, highlighting that open models are evolving yet not flawless;
Anthropic API Enhances Claude's Capabilities with New Web Search Tool Integration
• Anthropic's new API feature enables Claude to perform web searches, providing developers with real-time data and insights for more accurate and informed AI responses;
• This integration supports various industries, offering use cases from financial analysis to legal research by supplying up-to-date information from relevant and verified online sources;
• Organizations can tailor the AI's information sources through domain allow and block lists, ensuring that search results align with their unique content standards and privacy needs.
Enhanced Gemini 2.5 Pro Empowers Developers to Create Sophisticated Interactive Web Apps
• Early access to the updated Gemini 2.5 Pro Preview has been released ahead of Google I/O, featuring enhanced capabilities for building interactive web apps with improved coding functions;
• The Gemini 2.5 Pro now leads the WebDev Arena Leaderboard, earning +147 Elo points for its ability to create aesthetically pleasing and highly functional web apps with enhanced video understanding;
• The improved Gemini 2.5 Pro is available through Google AI Studio and Vertex AI, allowing users to build interactive web apps and use features like Canvas with a single prompt;
Google Releases Gemini 2.0 Flash Preview with Advanced Image Generation Features
• Gemini 2.0 Flash preview now includes image generation capabilities, available via the Gemini API in Google AI Studio and Vertex AI, offering developers enhanced integration options;
• The latest model update features improved visual quality, more precise text rendering, and significantly reduced filter block rates, enhancing the overall image generation experience compared to its experimental version;
• Developers express excitement over Gemini's ability to recontextualize products in new environments, showcasing impressive adaptability and creative potential in real-world applications;
Microsoft Enhances AI Systems with Open Agent Protocols for Cross-Platform Collaboration
• Over 10,000 organizations adopted Microsoft's new Agent Service in four months to deploy scalable agentic systems across enterprise environments
• Azure AI Foundry now supports open protocols like Agent2Agent (A2A), enabling diverse agents to collaborate across clouds, platforms, and organizational boundaries
• Microsoft Copilot is becoming a key interface for AI, facilitating interoperability as enterprises demand agents that orchestrate tasks spanning vendors and data silos.
Mistral Medium 3 Model Balances Cost and Top Performance for Enterprises
• Mistral Medium 3 sets new benchmarks in enterprise efficiency, offering state-of-the-art performance at eight times lower cost with seamless deployability to drive enterprise adoption;
• Capable of hybrid or on-premises deployment and integration into enterprise systems, Mistral Medium 3 excels in coding and multimodal understanding, surpassing many larger competitors;
• Available on platforms like Mistral La Plateforme and Amazon Sagemaker, Mistral Medium 3 provides a high-fidelity solution for complex enterprise needs, with imminent expansion to more cloud services.
Amazon's Vulcan Robot Uses Sense of Touch to Enhance Fulfillment Center Efficiency
• Amazon debuts Vulcan, a groundbreaking robot with a sense of touch, enhancing safety and efficiency at fulfillment centers by automating inventory tasks currently challenging for typical robots with no sensory feedback;
• Built on advances in robotics and physical AI, Vulcan enhances Amazon's operational efficiency by skillfully manipulating stored items and preventing accidental damage with its advanced force feedback system;
• Vulcan's deployment across Amazon facilities aims to improve workplace ergonomics and foster roles for employees to engage in more technical aspects of robotics alongside AI systems.
Figma Introduces Figma Make for Streamlined Prompt-to-Code Capabilities from Design
• Figma Make introduces a cutting-edge prompt-to-code feature, enabling designers to transform static designs into interactive prototypes using natural language prompts, streamlining the design-to-production workflow
• Figma Sites elevates design capabilities by allowing teams to publish designs directly to the web with a single click, supporting responsive layouts and custom code interactions for dynamic web experiences
• Figma Draw enhances visual design tools with an updated vector editing interface, offering features like multi-select vectors, pattern fills, and dynamic strokes to empower creative expression.
A Tsinghua University AI Virtual Hospital Aims to Broaden Healthcare Access Globally
• Tsinghua University will launch an AI-powered virtual hospital simulating 21 clinical departments, aiming to boost medical training and access in remote or conflict areas
• Trials by Tairex show AI doctors diagnose more efficiently, potentially easing China's stressed healthcare system by supporting doctors and reducing administrative tasks
• The platform employs a time-compression engine for rapid patient care simulation, designed to assist, not replace, doctors, improving access to quality care via smartphones;
OpenAI Plans $3 Billion Acquisition of AI Coding Tool Windsurf
• OpenAI is set to acquire the AI coding tool Windsurf for approximately $3 billion, aiming to enhance its offerings in AI-assisted programming tools if the deal closes;
• Windsurf, formerly known as Codeium, previously explored funding discussions with investors like Kleiner Perkins and General Catalyst, valuing the startup at $3 billion;
• With competitors like Anthropic and Microsoft’s GitHub already in the AI coding space, acquiring Windsurf could strengthen OpenAI's position in providing advanced code completion solutions;
Katy Perry's AI Images Steal the Spotlight at 2025 Met Gala
• AI-generated images of Katy Perry attending the Met Gala in a black latex gown went viral, fooling many into believing she was physically present at the event;
• Fans noticed minor distortions in the images, particularly around the waist, leading to the revelation that these hyper-realistic visuals were entirely AI-generated;
• Despite confirming the images were fake, fans praised the digital portrayal, claiming AI-Katy outshone real attendees, including stars like Kim Kardashian and Rihanna.
UNDP Report Highlights AI's Potential to Alleviate Stagnation in Human Development
• The 2025 Human Development Report by UNDP highlights unprecedented slowdowns in global development, with AI posited as a catalyst for reversing this trend
• The report emphasizes worsening global inequalities and stalled HDI progress, urging decisive actions against economic stagnation and advocating for AI-driven development paths
• Survey respondents indicate cautious optimism about AI's impact on employment and productivity, with a focus on human-centered AI integration across design, deployment, and modernized sectors;
Pinterest Expands Visual Search Capabilities to Enhance Personalized Fashion Discovery and Shopping
• Pinterest unveiled new visual search features, aiming to simplify how users discover and shop items that align with their personal style preferences through advanced image-based technology
• Users can now refine search results using a new customization bar, allowing for more precise content exploration by adjusting color, style, and occasion parameters
• The enhanced visual search functionality enables users to long-press on images, making it easier to investigate fashion items directly from their home feed and expand usage across multiple contexts.
Stripe Debuts AI-Powered Payment Model and Expands Stablecoin Financial Services Globally
• Stripe introduced the world's first AI foundation model for payments, enhancing transaction security by detecting card testing attacks and increasing performance across its payment suite
• Stripe expanded its money management capabilities with Stablecoin Financial Accounts, providing businesses in 101 countries the ability to hold stablecoins and conduct multi-currency transactions without FX fees
• A deeper partnership between Stripe and NVIDIA resulted in the fastest-ever migration to Stripe Billing, showcasing enhanced cloud gaming subscription processing capabilities in a record six weeks;
⚖️ AI Ethics
OpenAI Initiates International Collaborative Effort to Foster Democratic AI Development Globally
• OpenAI launched an initiative, OpenAI for Countries, to help nations develop AI infrastructure based on democratic principles, countering authoritarian approaches to AI development worldwide
• Collaborations will focus on building secure in-country data centers, customizing AI solutions, and enhancing safety controls for more robust, democratized AI deployment
• OpenAI aims to establish a national start-up fund to foster local AI ecosystems, with plans to partner with 10 nations or regions as part of its initial project phase.
OpenAI Establishes Data Residency in Asia for Enhanced Privacy Compliance
• OpenAI establishes data residency for ChatGPT Enterprise, ChatGPT Edu, and API Platform in Japan, India, Singapore, and South Korea, meeting regional data sovereignty requirements;
• Users can select data storage locations for enhanced control within API Platform and new ChatGPT workspaces, keeping customer content within specific Asian nations;
• OpenAI's data residency supports GDPR, CCPA compliance, with features like AES-256 encryption and a comprehensive Data Processing Addendum for enhanced data privacy and security.
Reddit Faces Backlash Over AI Bots, Plans New User Verification Measures
• Reddit experienced a scandal as researchers deployed AI bots in its “Change My View” subreddit, sparking concerns over authenticity and the platform's human-centric reputation;
• In response to AI impersonations, Reddit plans user verification measures via third parties, potentially affecting user anonymity on a platform traditionally favoring privacy and minimal personal data collection;
• The move toward verification aligns with a broader social media trend due to regulatory pressures, but critics warn of privacy risks and possible misuse of personal data by authorities.
Arizona Courtroom First: AI Brings Voice to Road Rage Victim at Sentencing
• In a groundbreaking use of AI, Chris Pelkey's family utilized the technology to create a video of him delivering a victim impact statement during his killer's sentencing in Arizona
• The AI-generated video featured Pelkey expressing forgiveness and sharing messages of love, marking a new application of AI in courtrooms, beyond its usual administrative uses
• Legal experts express concerns over potential misuse, warning that AI-generated evidence could become influential in trials, given its realistic portrayal and accessibility for creating deepfakes;
AI Image Recreation Trend Transforms Dwayne Johnson's Face into Abstract Art Forms
• AI's iterative image recreation trend distorts original photos with each repetition, resulting in humorously exaggerated or abstract transformations of images like Dwayne “The Rock” Johnson;
• Small deviations in AI image recreation compound over iterations, likened to a visual "telephone" game, leading to drastically altered results from the original image;
• Social media platforms like Reddit and TikTok are abuzz with AI-generated recursive images, showcasing both the technology's quirks and the embedded biases within AI models.
AI-Powered WhatsApp Scams Cost UK Families Nearly £500,000 by May 2025
• In the UK, AI-powered scams on WhatsApp have cost families nearly half a million pounds by May 2025, spotlighting the growing risk of technology-aided fraud
• Cybercriminals now use AI-generated voice messages to convincingly mimic loved ones in WhatsApp scams, combining old deception techniques with advanced generative AI technology
• Victims have reported 506 WhatsApp scams in 2025, losing £490,606, with the AI-enhanced "Hi Mum" scam relying on voice cloning for authenticity.
Former Facebook Insider Exposes Surveillance Capitalism's Exploitation of Vulnerable Youth for Profit
• Surveillance capitalism thrives as companies profit from user data by crafting hyper-personalized ads, leveraging consumer behavior metrics across social media platforms like Google, YouTube, and TikTok
• Recent exposés reveal alarming ad-targeting practices, with platforms like Meta allegedly exploiting psychological vulnerabilities among teens, raising ethical concerns in digital advertising
• Despite criticism, tech giants defend their targeting strategies, claiming aggregated data use, yet face ongoing scrutiny over the moral implications of monetizing user emotions and insecurities.
🎓AI Academia
New Framework Evaluates Chatbot Risks with Human-Centered Cybersecurity Metrics on GARAK
• A novel human-centered cybersecurity risk metric with a multidimensional perspective has been proposed to assess chatbots utilizing large language models, focusing on practical applicability
• The proposed metric has been successfully implemented in the open-source platform GARAK, enabling real-world application and testing of chatbot-related operational risks;
• Evaluations of this framework across different RAG-based system implementations show improved analysis, interpretability, and reproducibility of human-centric cybersecurity risks in chatbot environments.
Study Shows Humans Can Accurately Identify AI Texts With Immediate Feedback Training
• A recent study demonstrates that humans can improve their ability to distinguish between human and AI-generated texts when provided with immediate feedback;
• Feedback significantly enhances the accuracy and confidence of individuals in identifying AI-generated texts, correcting initial misconceptions about AI text traits such as stylistic rigidity and readability;
• The research underscores the importance of feedback in educational settings to better equip individuals in identifying AI-generated content, highlighting the risks of false confidence without proper training.
Generative AI Poised to Transform Key Aspects of Transportation Planning Frameworks
• Researchers surveyed generative AI's transformative potential in transportation planning, highlighting applications in demand forecasting, infrastructure design, policy evaluation, and traffic simulation;
• A new taxonomy was proposed to classify generative AI applications and methods, focusing on transportation planning tasks and computational techniques for advancing mobility systems;
• Key challenges addressed include data scarcity, bias mitigation, and developing evaluation frameworks that align AI with transportation objectives like sustainability, equity, and system efficiency.
Study Evaluates AI's Ethical Risk and Bias Detection Using Role-Playing Techniques
• Researchers have applied the DOSPERT scale to assess the risk‑taking tendencies of Large Language Models across various domains, with a strong emphasis on ethical decision-making;
• A novel Ethical Decision-Making Risk Attitude Scale (EDRAS) has been introduced to quantify ethical biases and risk attitudes in LLMs through role-play assessments;
• The study reveals systematic biases in LLMs, enhancing understanding of their risk personalities and aiding in the development of fairer and more trustworthy AI systems.
Research Highlights Security Challenges to Large Language Models Amid Rising Attacks
• A recent paper highlights vulnerabilities in Large Language Models (LLMs) to prompt injection and jailbreaking attacks, necessitating robust defense strategies for safe deployment
• Researchers have categorized LLM attacks into prompt-based, model-based, multimodal, and multilingual types, emphasizing the need for adaptive countermeasures
• The study evaluates defense mechanisms, such as prompt filtering and multi-agent defenses, while highlighting the importance of addressing ethical and societal impacts in LLM security.
Study Highlights Accountability Patterns in AI Incident Response and Societal Learning
• A recent preprint examines the limitations and insights of the AI Incident Database, showing it falls short in documenting AI implementation failures due to media reliance;
• Analysis of 962 incidents reveals patterns of responsibility and accountability, indicating social context and controversy shape responses more than mere identification of responsible parties;
• Highlighted in the study, AIID's true value lies in its ability to document societal reactions to AI failures, offering insights into adaptive governance and social learning.
European Social Scoring Ban in AI Act: Effective Safeguard or Insufficient Measure?
• The AI Act bans social scoring practices inspired by China's Social Credit System, but allows European scoring in areas like creditworthiness and employee monitoring under high-risk AI requirements
• Despite seeming ineffectual, the ban on social scoring is designed as a flexible safeguard to prevent authoritarian application in Europe, complementing GDPR's protections
• The AI Act aims to balance high-risk AI system authorization and individual protection against disproportionate AI scoring, aligning with GDPR and promoting fair use of automated decision-making.
About SoRAI: SoRAI is committed to advancing AI literacy through practical, accessible, and high-quality education. Our programs emphasize responsible AI use, equipping learners with the skills to anticipate and mitigate risks effectively. Our flagship AIGP certification courses, built on real-world experience, drive AI governance education with innovative, human-centric approaches, laying the foundation for quantifying AI governance literacy. Subscribe to our free newsletter to stay ahead of the AI Governance curve.
