Senator Bernie Sanders raises concerns over AI 'doomsday scenarios'
Last Week in AI Policy #26 - July 21, 2025
Delivered to your inbox every Monday, Last Week in AI Policy is a rundown of the previous week’s happenings in AI governance in America. News, articles, opinion pieces, and more, to bring you up to speed for the week ahead.
Policy
Sen. Bernie Sanders (I-VT) talks job losses and existential risks from AI.
Citing top researchers, Sanders voices concern that humanity could lose control of advanced AI systems, a so-called “doomsday scenario.”
Sen. Sanders says AI’s productivity windfall risks enriching the “billionaire class.”
He foresees “massive job losses” and warns bosses could threaten to swap humans for machines to extract wage concessions.
He argues workers should capture those gains through measures such as a 32‑hour work week with no pay cut.
Several lawmakers raised concerns over antisemitic posts from Elon Musk’s Grok AI
A July 11 letter from Reps. Josh Gottheimer (D‑NJ), Tom Suozzi (D‑NY), and Don Bacon (R‑NE) urges Elon Musk to address extremist and violent content produced by xAI’s Grok chatbot.
They recount Grok posts on X praising Adolf Hitler, endorsing Nazi ideology, and issuing graphic rape fantasies and instructions.
The lawmakers label the episode a profound breach of public trust, press for immediate deletion of the posts, and warn of harm to children and other users.
Their letter demands detailed explanations of Musk’s July 4 “improvements,” Grok’s training data and filters, and the safeguards xAI will implement to prevent future extremist outputs.
Twenty additional House members from both parties signed on, suggesting broad congressional scrutiny of AI safety and content moderation.
At the 15 July 2025 Energy and Innovation Summit in Pennsylvania, President Trump announced more than $90 billion in combined artificial intelligence and energy projects.
Google committed $25 billion to new data center and AI infrastructure, Blackstone earmarked $25 billion for data centers plus natural gas plants, and CoreWeave slated $6 billion for more GPUs and facility expansion.
The White House framed the package as cementing U.S. dominance in AI while driving job creation and infrastructure investment.
The Pentagon’s Chief Digital and Artificial Intelligence Office issued individual contracts (each worth up to $200 million) to Anthropic, Google, OpenAI, and xAI to furnish “advanced AI capabilities” for national-security tasks.
Chief Digital and AI Officer Doug Matty said commercial solutions will accelerate AI adoption across war-fighting, intelligence, business, and enterprise information systems, preserving U.S. strategic edge.
At the same time, Elon Musk’s xAI unveiled “Grok for Government,” adding its models to the GSA schedule so any federal agency can buy them.
In a July 17 address to the NBER Summer Institute, Federal Reserve Governor Lisa D. Cook said the “incredibly rapid” diffusion of AI is accelerating idea generation, raising worker efficiency, and will shape Fed objectives.
Cook emphasized that the Fed already uses large‑language models for drafting, coding, and economic analysis but “does not” rely on them when setting interest rate policy.
She also forecast that AI‑driven productivity gains could eventually lower inflation, though an initial investment surge may lift prices, and she warned of job reallocation as tasks are automated.
Internal research presented in the speech shows LLMs can accurately interpret Fed minutes, while separate machine‑learning work aims to improve financial crisis forecasting - evidence Cook cited to justify continued experimentation.
Nearly half of the GSA is using their internal AI chatbot daily
Acting Administrator Stephen Ehikian told the Government Efficiency Summit that roughly 50 percent of GSA’s workforce uses the internal generative AI assistant “GSAi” every day.
Introduced in March, GSAi is already drafting property listings, reviewing contracts, suggesting outcome‑based pricing, and tutoring staff in new programming languages; employees submit fresh use case requests daily.
Press Clips
Aggressive timeline prediction for artificial superintelligence by Rep. Andy Biggs (R-AZ), putting it before 2030 (Peter Wildeford) 🐦
An exchange from Rep. Scott Perry at a recent US House hearing about AI (Daniel Eth) 🐦
After the ChatGPT Moment: Measuring AI’s Adoption (Epoch AI) ✍
Datacenter Delusions: Simply building compute capacity does not make for a convincing national AI strategy (Anton Leicht) ✍
Memo to the President: A Conversation with Jensen Huang (Special Competitive Studies Project) 📽
The struggle over AI in journalism is escalating (Blood in the Machine) ✍
The Next Modest Chapter in the TSMC Chronicles (Semiconductor Business Intelligence) ✍
Import AI 420: Prisoner Dilemma AI; FrontierMath Tier 4; and how to regulate AI companies (Import AI) ✍