AI Week in Review - 3/8/25

Public Sector

Author

Chris Kraft
March 08, 2025

Good day, public sector AI enthusiasts! GenAI models continue to improve with multiple providers now offering enhanced research capabilities. What should we expect from providers going forward? There’s speculation that OpenAI will be offering PhD-level AI capabilities for thousands per month.

If you’re having a hard time staying caught up on the latest models, you should check out Andrej Karpathy’s recent video on how he uses large language models. There’s also some great training from Hugging Face on agents and reasoning models. And if that’s not enough - Stanford just released updated videos from Chris Manning’s legendary NLP course CS224N.

In this week’s edition:

  • PhD-level AI?

  • OSTP AI Action Plan

  • How Does Andrej Karpathy use LLMs?

  • The latest public (and private) sector news

The Highlights

Source: ChatGPT

Rumor has it that OpenAI will be introducing a premium “PhD-level AI” plan at $20,000 per month. This advanced subscription targets enterprises and research-driven users looking for powerful analysis, larger context windows, and more autonomous AI agents than standard offerings.

More Details

  • Deep AI reasoning: Capable of doctoral-level tasks, including technical research and high-level data analysis.

  • Expanded context window: Handles more information, ideal for lengthy documents and data-heavy projects.

  •  Autonomous agents: Supports iterative tasks with minimal human involvement—analyzing data, drafting reports, and proposing next steps.

  • Enterprise tier pricing: $20,000 monthly subscription, aimed at organizations that need serious AI power and can justify the cost with big productivity gains.

Why It Matters
A “PhD-level” AI could reshape how businesses tackle R&D, knowledge management, and advanced analytics. By rapidly processing and synthesizing massive data sets, it promises near-instant insights—potentially freeing knowledge workers to focus on higher-level strategy and innovation.

My Visit to OSTP

​The White House recently issued an Request for Information (RFI) seeking public input to develop an Artificial Intelligence (AI) Action Plan

More Details

Why It Matters
There’s still time left to provide comments, but not much! Make sure you get your comments in before the March 15th deadline.

How Does Andrej Karpathy use LLMs?

In this video, AI legend Andrej Karpathy shares how he uses Large Language Models (LLMs) in his life.

More Details
This 2-hour video dives into a broad range of topics, including:

  • ChatGPT Mechanics: Hits on the internal workings of ChatGPT, explaining how it processes and generates responses.

  • Basic Interaction: Highlights basic ways to interact with LLMs, highlighting their flexibility with various prompts.​

  • Advanced Thinking Models: Explores more sophisticated models capable of deeper reasoning.​

  • Tool Integration: Demonstrates how LLMs can be integrated with tools like internet search and code interpreters.​

  • Data Analysis and Visualization: Advanced data analysis, including generating figures and plots.​

Most of Andrej’s videos are very technical and they provide amazing insight into AI technologies. This video is a great treat for a broader audience to benefit from a less technical dive into LLMs.

The Details

Federal

Democracy Forward requests DOGE AI info
Democracy Forward has filed FOIA requests to investigate the Trump administration’s reported use of AI in federal personnel decisions, particularly in employee performance reviews.

AI to help regulators prevent unsafe products
A House bill from Rep. Darren Soto directs the Consumer Product Safety Commission to pilot AI for tracking product hazards, recalled items, and consumer safety trends.

AI-driven wargaming accelerates military mission planning
Johns Hopkins APL is developing AI-powered wargaming tools that automate scenario design and analysis, dramatically reducing planning time and enhancing strategic decision-making for military and government agencies.

DOGE all-in on AI
Elon Musk’s Department of Government Efficiency is advocating AI to streamline federal operations, but secrecy around its implementation raises concerns about oversight, bias, and public trust.

Pentagon signs AI deal to help commanders plan military maneuvers
The U.S. military’s new "Thunderforge" initiative will use AI from Scale AI, Microsoft, and Google to enhance decision-making, integrating real-time intelligence for strategic maneuver planning.

Asymmetric AI warfare in the era of cheap intelligence
As AI reshapes asymmetric warfare, the U.S. must balance innovation, security, and ethics through strategic investments, governance, and public-private collaboration to maintain global leadership.

State Dept. to use AI to revoke visas of foreign students who appear "pro-Hamas"
The State Department is deploying AI to review social media activity and revoke visas of foreign students suspected of supporting Hamas, intensifying government oversight of free speech and national security threats.

Anthropic’s recommendations to OSTP for the U.S. AI Action Plan
Anthropic's AI policy recommendations to OSTP focus on national security, export controls, lab security, energy infrastructure, government AI adoption, and economic preparedness to maintain U.S. leadership in AI development.

Trump and Taiwanese AI chipmaker announce new $100 billion plan
TSMC and Trump announced a $100 billion investment to build five new U.S. chip factories, strengthening domestic semiconductor production and reducing reliance on Asian supply chains.

Army deploys AI tool to purge DEIA content from training materials
The U.S. Army is using an AI tool called CamoGPT to review and remove diversity, equity, inclusion, and accessibility (DEIA) references in training materials, aligning with President Trump’s executive order.

Trump puts his stamp on AI
Trump is reshaping U.S. AI policy by revoking Biden-era safeguards, backing industry-led initiatives like Stargate, and prioritizing minimal regulation to maintain U.S. dominance in the global AI race.

Voters don’t want government to make decisions using AI
A new poll finds that 55% of U.S. voters oppose government use of AI for decision-making, citing concerns over discrimination and misuse, while nearly 60% worry about insufficient AI regulation.

Commentary: AI censorship and green computing could weaken the West in AI race against China
A new commentary warns that Western AI regulations, green computing policies, and ideological biases risk ceding technological ground to China, where AI is tightly controlled to serve authoritarian interests.

AI is key to Trump’s education overhaul
While the Trump administration pushes to reduce federal oversight in education, experts argue that fostering AI adoption is crucial to ensuring innovation and personalized learning opportunities for all students.

Civil Society lobbies to keep the AI use case inventories
Eighteen organizations are calling on the Trump administration to keep federal AI use case inventories, emphasizing their role in transparency, trust, and innovation across government agencies.

State / Local

Penn State begins ‘cheat-a-thon’ to test GenAI
Penn State’s "cheat-a-thon" challenges students to test generative AI’s limits by answering tough faculty-designed questions using AI alone, aiming to explore its benefits and risks.

Iowa pushes nuclear energy plan to support AI
Iowa is advancing legislation and a Nuclear Energy Task Force to support growing AI and data center energy demands, including efforts to recommission the Duane Arnold nuclear plant.

Is NJ’s power grid prepared for large-scale AI data centers?
New Jersey lawmakers are assessing whether the state’s power grid can handle AI-driven energy demands, as data center electricity consumption is projected to triple within five years.

State lawmakers weigh in on health care AI
State lawmakers are increasingly introducing bills to regulate AI in health care, focusing on transparency, payer use, and clinical decision-making to balance innovation with patient protection.

Keeping patient care human—Connecticut fights AI-driven denials
A Connecticut bill proposes prohibiting health insurers from using AI to determine patient care, aiming to protect access to necessary treatments and prevent automated denials.

California labor unions take on AI surveillance in the workplace
The California Labor Federation is backing new bills to limit AI-powered employee monitoring, aiming to protect worker privacy and set a national precedent for workplace surveillance regulations.

Arkansas launches AI-powered job portal to streamline hiring
Arkansas' new LAUNCH platform uses AI and data-driven matching to connect job seekers with employers, integrating career resources and training opportunities for a more efficient workforce system.

The impacts of a fragmented AI legislative landscape
With no federal AI legislation, states are crafting their own policies, creating a fragmented but evolving regulatory landscape that fosters learning, innovation, and governance clarity.

Oklahoma’s new AI watches procurement for filing errors
Oklahoma’s Office of Management Enterprise Services is using AI-powered Process Copilot to reduce procurement errors, enhance oversight, and improve efficiency in managing taxpayer funds.

CT leaders push AI education and policy
Connecticut lawmakers and business leaders are expanding AI education and workforce training to ensure residents benefit from AI's economic potential while preparing for regulatory challenges.

Yale students shape Connecticut’s AI policy
Yale students advised Connecticut Gov. Ned Lamont on AI policy, emphasizing sector-specific regulations and industry collaboration, leading to proposed legislation balancing innovation with risk management.

Arizona explores AI to boost disaster prediction and emergency response
Arizona emergency officials and ASU researchers are developing an AI-powered system to enhance disaster prediction and response, aiming to improve decision-making without replacing human roles.

AI workers may get special protections in Colorado
Colorado lawmakers are debating a bill to grant whistleblower protections to AI workers reporting public safety risks, sparking concerns from business groups and Gov. Polis over potential economic impacts.

Minnesota moves to ban AI ‘nudify’ apps
Minnesota lawmakers are considering a bill to ban AI-powered "nudification" apps, aiming to prevent the creation of explicit deepfake images, though legal experts warn of potential First Amendment challenges.

AI-powered cameras help New Orleans cut crime by 50%
AI-assisted surveillance cameras and increased state police presence have helped New Orleans significantly reduce crime, contributing to a 50% drop in homicides and armed robberies.

Nevada moves to outlaw AI-generated child exploitation
Nevada lawmakers are proposing bipartisan legislation to criminalize AI-generated child pornography, updating laws to address emerging online threats to children’s safety.

New Hampshire leverages AI to modernize benefits and streamline unemployment
New Hampshire is integrating AI into its already modernized benefits system to streamline unemployment claims, improve efficiency, and enhance fraud detection, building on past digital transformation efforts.

NJ weighs energy challenges as it pushes to lead in AI innovation
New Jersey lawmakers are exploring how to meet the growing energy demands of AI and data centers amid a declining energy supply, considering renewables, natural gas, and incentives to stay competitive.

AI tells New York subway workers suspicious sound is a problem
The MTA is piloting AI-powered sensors on subway cars to detect track issues, improving maintenance efficiency, reducing delays, and enhancing safety for workers and riders.

Cornell receives $10.5M gift to boost AI research through Empire AI
A $10.5 million gift from Bloomberg co-founder Tom Secunda will fund AI research at Cornell, supporting postdoctoral fellowships, a public AI conference, and access to Empire AI’s computing infrastructure.

Oregon moves to ban AI from using the title 'nurse' to protect patient trust
The Oregon House passed a bill ensuring only human practitioners can use the title "nurse," preventing AI systems from misleading patients and protecting the integrity of the profession.

AI’s rising water demands pose new challenges for drought-stricken Utah
AI's growing reliance on water-intensive cooling systems raises concerns in drought-prone Utah, where expanding data centers could further strain the state's already limited water resources.

Kentucky Senate advances AI regs to ensure transparency and election integrity
The Kentucky Senate passed a bill establishing AI oversight in state government, requiring transparency, accountability, and safeguards against AI-generated election misinformation.

Virginia turns to AI to cut road costs
Facing rising construction expenses, VDOT is piloting AI-driven solutions for cost estimation and pavement management to optimize investments and maintain Virginia’s vast road network.

Could a pause on building data centers in Va. benefit N.C.?
As Virginia considers restricting data center expansion, North Carolina’s existing infrastructure and tax incentives could position it as a growing hub for hyperscale data center investments.

International

South Korea’s AI policies under scrutiny amid concerns over bias and surveillance
A new report examines AI policy and risks in South Korea’s public sector, law enforcement, education, and social welfare, highlighting concerns over bias, surveillance, and inadequate regulation.

Malaysia launches AI adaptation guidelines for public sector
Malaysia has launched AI adaptation guidelines for the public sector, emphasizing responsible use, transparency, and risk management to ensure AI-assisted decisions serve the public interest.

Top 10 African countries leading in AI usage and development
South Africa, Nigeria, and Kenya are leading AI adoption in Africa, with strong government support, research initiatives, and AI-driven solutions transforming industries like fintech, healthcare, and agriculture.

China tells its AI leaders to avoid US travel over security concerns
China is reportedly advising top AI researchers and entrepreneurs to avoid U.S. travel over concerns about information leaks and potential detentions amid rising geopolitical tensions.

Argentina as the world’s next AI hub?
Argentina aims to position itself as a global AI hub by leveraging cheap energy, skilled talent, and deregulation, but experts question its ability to compete with established tech giants.

Taiwan to expand AI use at customs clearance points
Taiwan plans to enhance customs security by expanding AI-based X-ray imaging to detect contraband, including meat, e-cigarettes, and banknotes, with a NT$275.77 million investment through 2028.

Hanoi deploys AI hotline
Hanoi has launched an AI-powered hotline (19001009) to streamline public services, offering 24/7 multi-channel support for administrative inquiries via phone, website, chatbot, and mobile app.

Canada launches first-ever AI Strategy for the federal public service
Canada’s federal government launches its first AI Strategy, focusing on responsible AI use, workforce training, enhanced digital services, and transparency to improve public service efficiency.

Use of AI in the Judiciary
A new report highlights 125 AI-driven cyberjustice tools enhancing judicial efficiency, accessibility, and transparency, while emphasizing the need for oversight and responsible integration in legal systems.

China says it will increase support for AI, science and tech innovation
China’s government pledges increased support for AI model applications, venture capital, and next-gen technologies like quantum computing and 6G to foster innovation and self-sufficiency.

China doubles down on AI and tech innovation as Trump ratchets up trade pressure
Facing new U.S. trade restrictions, China is doubling down on AI, quantum tech, and 6G innovation, prioritizing self-reliance and private sector support to counter external economic pressures.

Russian disinformation infiltrating AI chatbots
A NewsGuard study found that Russian disinformation from the Pravda network has infiltrated AI chatbots, with leading models repeating pro-Kremlin falsehoods 33% of the time, raising concerns about AI-driven misinformation.

AI in Qatar: Assessing the potential economic impacts
Qatar's AI adoption is accelerating, with policy reforms and private sector contributions driving productivity gains, positioning the country as a leader among emerging markets.

Competency Framework for AI Integration in India
IndiaAI launched a Competency Framework to enhance AI literacy among public sector officials, addressing skill gaps and fostering AI-driven efficiency, innovation, and governance in government services.

UK launches £5M AI Security Fund
The AI Security Institute's new Challenge Fund offers grants of up to £200,000 for research tackling AI security threats, aiming to enhance oversight, prevent misuse, and boost public trust in AI adoption.

DeepSeek, Huawei, export controls, and the future of the U.S.-China AI race
DeepSeek’s rapid rise challenges U.S. dominance in AI, highlighting risks from chip export loopholes, IP theft, and China’s push for semiconductor self-sufficiency amid tightening U.S. controls.

Everything Else

AI’s role in preventing the next pandemic
AI-driven modeling and predictive analytics are revolutionizing global health security, enabling scientists to anticipate outbreaks, assess risks, and create early warning systems to prevent future pandemics.

GM hires first-ever AI chief
General Motors has appointed Barak Turovsky as its first AI chief, tasking him with overseeing AI strategy across autonomous vehicles, logistics, manufacturing, and customer experience.

Anthropic raises $3.5B at $61.5B valuation
Anthropic has raised $3.5 billion at a $61.5 billion valuation to advance AI research, expand globally, and enhance Claude’s capabilities in coding, interpretability, and enterprise applications.

Here’s what happens when AI takes over performance reviews
AI-powered performance reviews can improve efficiency and standardization, but experts warn against over-reliance, as data quality issues and lack of human nuance could worsen existing biases.

How the risk of AI weapons could spiral out of control
From flawed medical diagnoses to AI-powered weapons, researchers and policymakers grapple with the risks of AI "thinking" in unpredictable ways, raising urgent ethical and security concerns.

How AI tools are reshaping the coding workforce
GenAI coding tools like GitHub Copilot are streamlining software development, enabling leaner teams, shifting developer priorities to AI prompting, and raising hiring expectations across the tech industry.

AWS bets big on agentic AI
Amazon’s AWS has formed a new agentic AI group, led by Swami Sivasubramanian, to drive automation and innovation, positioning AI agents as the next multi-billion-dollar business.

The McDonald’s AI makeover
McDonald’s is rolling out AI-powered drive-throughs, predictive maintenance, and edge computing to improve order accuracy, streamline operations, and enhance customer and employee experiences.

Llama 4 -here come the AI agents
Meta’s upcoming Llama 4 AI model will enhance AI agents with reasoning capabilities, enabling automation of complex business tasks and real-time web interaction.

Google expands AI-powered search, sparking competition and concerns
Google is integrating its Gemini 2.0 AI model into search, enhancing AI-generated overviews, but critics worry about its impact on web traffic, misinformation, and monopoly concerns.

LA Times pulls AI tool after criticism over downplaying KKK history
The LA Times removed its new AI-driven “Insights” tool from an article after it downplayed the KKK’s history, sparking controversy and concerns from journalists about AI-generated analysis.

OpenAI launches $50M NextGenAI initiative to drive academic AI innovation
OpenAI’s $50M NextGenAI initiative partners with 15 top institutions, including Oxford, Harvard, and MIT, to integrate AI into education, accelerate scientific research, and expand access to AI-powered tools.

OpenAI’s “PhD-Level AI” Sparks Debate Over Cost, Capability, and Research Value
OpenAI’s rumored $20,000/month AI agents claim to match PhD-level expertise, but questions remain about their real-world research value versus human intellectual capabilities.

Apple delays Siri improvements – pushing into 2026
Apple has postponed Siri’s advanced AI features until 2026, delaying app integration and personalization improvements as rivals like OpenAI, Google, and Amazon push forward with AI assistants.

Putting explainable AI to the test
A literature review of AI-enabled recommendation systems reveals inconsistencies in explainability evaluations, suggesting policymakers should invest in standardized AI safety assessments and workforce expertise.