AI News Roundup: OpenAI Proposes Robot Taxes & 4-Day Workweek, Google Drops Gemma 4, NVIDIA Kicks Off Robotics Week

OpenAI Calls for Robot Taxes, a Public Wealth Fund, and a Four-Day Workweek

OpenAI released a 13-page policy paper titled “Industrial Policy for the Intelligence Age” this week, laying out what CEO Sam Altman describes as a “new social contract” on the scale of the New Deal. The document argues that AI superintelligence is close enough — and disruptive enough — that America needs to fundamentally rethink how it taxes, regulates, and redistributes wealth.

The headline proposals are striking: a robot tax modeled on the idea that automated systems should pay the same taxes as the workers they replace, a nationally managed public wealth fund seeded by AI companies to give every American a stake in AI-driven growth, and government-incentivized pilots of a 32-hour workweek at full pay — converting AI-driven efficiency into time back for workers. OpenAI also proposes shifting the tax burden from labor to capital, with higher rates on corporate income and AI-driven returns.

The timing is notable: OpenAI is simultaneously preparing for a potential IPO later this year while positioning itself as a responsible steward of transformative technology. Critics have been quick to point out the tension between calling for higher corporate taxes while seeking a sky-high public valuation.

Google Releases Gemma 4: Open Models That Punch Way Above Their Weight

Google DeepMind released Gemma 4 on April 2, its most capable open model family yet, and the AI community has been putting it through its paces all week. The standout: Gemma 4 is now ranked #3 on Arena AI’s open model leaderboard, beating models 20 times its size. The family includes four variants ranging from a 2B model that runs on smartphones to a 31B dense model for cloud deployment.

Perhaps more significant than the benchmarks is the licensing shift. Google abandoned its restrictive custom license in favor of Apache 2.0 — the same permissive terms used by Qwen, Mistral, and the broader open-source community. With context windows up to 256K tokens, native vision and audio processing, and fluency in over 140 languages, Gemma 4 represents a major democratization play. Developers can now run state-of-the-art agentic AI on everything from a Raspberry Pi to a consumer GPU.

NVIDIA Kicks Off National Robotics Week with Physical AI Showcase

It’s National Robotics Week, and NVIDIA is using the occasion to highlight how rapidly AI is crossing from the digital world into the physical one. The company showcased breakthroughs across agriculture, manufacturing, energy, and logistics — all powered by its accelerated computing platforms, Omniverse simulation tools, and the Isaac Sim robotics framework.

The real-world results are impressive: Maximo, a solar robotics company, completed a 100-megawatt solar installation using its autonomous robot fleet built on NVIDIA’s stack. Aigen’s solar-powered robots are using vision AI for precision weed control, eliminating the need for chemical herbicides. Coming off GTC 2026 in San Jose — which featured 110+ robots on the show floor and partnerships with ABB, FANUC, Figure AI, and Boston Dynamics — physical AI is clearly moving from demos to deployments.

OpenAI’s IPO Preparations Hit Internal Turbulence

Even as OpenAI surpasses $25 billion in annualized revenue and prepares for a potential public listing, internal friction is complicating the timeline. CFO Sarah Friar has reportedly questioned the company’s IPO readiness, citing concerns about revenue growth sustainability and ballooning capital expenditures on AI infrastructure. CEO Sam Altman, by contrast, continues to push for a 2026 listing.

The strategic tension reflects a broader industry challenge: AI companies are generating enormous top-line revenue but spending even more on compute. OpenAI has already made six acquisitions in 2026 — nearly matching all of 2025 — most recently open-source developer tools maker Astral. Whether the public markets will reward growth-at-all-costs or demand a clearer path to profitability remains the $100 billion question.

California’s Newsom Signs AI Procurement Executive Order

Governor Gavin Newsom signed executive order N-5-26 requiring companies that want to do business with the California state government to disclose their AI usage and policies around bias, civil rights, and illegal content distribution. The order also mandates watermarking of AI-generated media in government contracts and directs state agencies to accelerate their own adoption of AI tools — with guardrails.

The move is deliberately positioned as a counterpoint to the Trump administration’s lighter-touch approach. With over 600 AI bills introduced across state legislatures in 2026 and a federal preemption fight brewing, California — home to nearly every major AI lab — is setting up to be the de facto regulatory proving ground. Newsom, widely seen as a 2028 presidential contender, is betting that “responsible AI” will be a winning platform.

Reality Check: Only 28% of Enterprise AI Projects Deliver ROI

A new Gartner report pours cold water on the AI hype cycle, revealing that just 28% of AI projects in infrastructure and operations deliver meaningful returns, while 20% fail outright. The finding comes even as enterprise adoption accelerates — Gartner projects that 40% of business applications will employ AI agents by end of 2026, up from under 5% in 2025.

The disconnect highlights a familiar pattern in enterprise tech: rapid adoption driven by FOMO, followed by a painful reckoning on integration costs, data quality, and organizational readiness. On the bright side, Automation Anywhere reported that AI agents now auto-resolve over 80% of IT support requests at companies that have invested in proper implementation, cutting service management costs by up to 50%.