Top Stories
OpenAI Wants Robot Taxes, a Public Wealth Fund, and a Four-Day Workweek
OpenAI published a 13-page policy document Monday titled “Industrial Policy for the Intelligence Age: Ideas to Keep People First” — and it reads like a company trying to get ahead of the political backlash it knows is coming.
The headline proposals: a nationally managed public wealth fund seeded by AI company contributions, modeled after Alaska’s Permanent Fund, that would invest in AI firms and businesses adopting the technology and distribute returns directly to citizens. Taxes on automated labor, with each robot paying what the human it replaced would have owed. Pilot programs for a 32-hour workweek with no loss in pay. And a shift in the tax base from payroll toward capital gains and corporate income, acknowledging that AI-driven growth could hollow out the funding for Social Security, Medicaid, SNAP, and housing assistance.
The document also proposes “containment playbooks” for autonomous, self-replicating AI systems and automatic safety net triggers that kick in when AI displacement metrics hit preset thresholds. CEO Sam Altman told Axios that cyberattacks via near-future AI models are “totally possible” within a year, and that using AI to create novel pathogens is “no longer theoretical.”
The timing is calculated. OpenAI is preparing for an IPO (potentially Q4 2026), recently closed a $122 billion funding round, and Congress is actively debating AI legislation. Publishing a policy blueprint that calls for taxing the very automation you’re building is a pre-emptive move — positioning yourself as a responsible actor before regulators decide you’re a target. Whether any of these proposals survive contact with actual policymaking is another question.
Source: TechCrunch, The Next Web, The Hill
Tufts Researchers Cut AI Energy Use by 100x With Neuro-Symbolic Approach
A team at Tufts University has published results that could reshape how we think about AI’s energy appetite — if they hold up at scale. By combining traditional neural networks with symbolic reasoning, Matthias Scheutz’s lab built a system that uses 1% of the energy required for training and 5% during operation compared to standard approaches.
The method, called neuro-symbolic AI, pairs the pattern-recognition strengths of neural networks with the rule-based logic that humans use to break problems into steps. Instead of brute-forcing solutions through trial and error, the system applies constraints that dramatically narrow the search space.
The numbers are striking. On a standard Tower of Hanoi benchmark, the hybrid system hit 95% accuracy versus 34% for conventional visual-language-action models. On more complex variants the system hadn’t seen during training, it succeeded 78% of the time while standard models failed every attempt. Training took 34 minutes instead of over 36 hours.
The catch: this work focuses on robotics tasks, not large language models. The Tower of Hanoi is a structured puzzle, not the messy, open-ended problems that LLMs tackle. Whether neuro-symbolic approaches can deliver similar gains for the models consuming the most energy — the trillion-parameter generative systems behind ChatGPT and Claude — remains an open research question. But as the industry faces a $7 trillion infrastructure bill and growing scrutiny over power consumption, any path to 100x efficiency gains deserves serious attention. The work will be presented at the International Conference of Robotics and Automation in Vienna this May.
Source: ScienceDaily, Tufts Now, SciTechDaily
Meta Cuts 200 More Jobs as AI Restructuring Continues
Meta is laying off roughly 200 employees in its Bay Area offices, with 124 positions cut in Burlingame and 74 in Sunnyvale, according to filings with California’s Employment Development Department. The cuts are permanent and scheduled for late May.
This follows a 1,500-person reduction at Reality Labs earlier this year and additional cuts across recruiting, sales, and operations in March. The pattern is consistent: traditional middle management is being replaced by “AI-first” technical roles — AI Builders and Pod Leads who are expected to be hands-on technical contributors rather than people managers.
The restructuring is happening alongside a projected $115–135 billion in capital expenditure for 2026, most of it directed at AI infrastructure. Meta is simultaneously spending more money on AI than almost any company in history while cutting the humans who built its previous products. The company frames it as realignment. Critics call it a convenient narrative that lets you fire people while your stock price goes up. Either way, Meta joins Oracle (30,000), Dell, and a growing list of tech companies that have cut over 61,000 AI-attributed jobs in 2026 alone.
Source: The Tech Portal, Fox Business
Quick Hits
-
California’s AI procurement order takes shape: Governor Newsom’s Executive Order N-5-26, signed March 30, requires AI vendors seeking state contracts to disclose their policies on illegal content distribution, bias, and civil rights protections. The order also creates the nation’s first state-level watermarking standards for AI-generated images and video, and notably allows California to override federal supply-chain risk designations for AI vendors — widely read as a response to the Pentagon-Anthropic procurement dispute. Vendor certification recommendations are due within 120 days. Governor’s Office
-
New York trains 100,000 state workers on AI: Governor Hochul expanded the AI Pro training tool — powered by Google Gemini — to all 100,000+ New York state employees across 50 agencies, making it the largest state workforce AI training program in the country. A pilot with 1,200 participants across eight agencies ran last fall. The two-part program covers responsible AI use for public sector employees. New York has also committed $500 million to the Empire AI computing center at SUNY Buffalo. Governor’s Office
-
OpenAI launches Safety Fellowship for outside researchers: A new five-month fellowship (September 2026 – February 2027) offers external researchers $3,850 per week — over $200,000 annualized — to work on AI safety and alignment. Priority areas include safety evaluation, robustness, agentic oversight, and high-severity misuse domains. Fellows get API credits and mentorship but no internal system access. Workspace is available at Constellation in Berkeley. Applications close May 3. OpenAI
-
AI-driven layoffs cross 61,000 in 2026: The tech sector reported 18,720 job cuts in March alone, up 24% year-over-year, with over 52,000 total tech cuts in Q1. At least 45 CEOs have explicitly cited AI efficiencies as the reason for headcount reductions. Goldman Sachs warned displaced workers to expect extended job searches and lower pay at their next role. The irony: companies are spending record amounts on AI while arguing they can’t afford to keep the people who built the products that fund the AI investment. Bloomberg
Worth Watching
OpenAI’s policy blueprint is worth reading in full, not because the proposals are likely to become law, but because of what it signals about the industry’s own expectations. When the company building the technology warns that AI could hollow out the tax base, displace enough workers to require automatic safety net triggers, and create self-replicating systems that need “containment playbooks,” that’s not an advocacy document — it’s a risk disclosure dressed as policy. The question is whether anyone in Congress treats it that way, or whether the pre-IPO positioning works as intended. Meanwhile, California and New York are doing what states do when the federal government stalls: filling the vacuum with their own rules. Between Newsom’s procurement standards and Hochul’s workforce training, the state-level AI governance framework is forming faster than most people realize.