Top Stories
White House Unveils AI Policy Framework, Wants Federal Rules to Override State Laws
The Trump administration released its national AI policy framework on Friday, laying out recommendations for Congress that would establish federal oversight while preempting state-level AI regulations.
The framework outlines six guiding principles: protecting children, preventing electricity costs from surging, respecting intellectual property rights, preventing censorship, educating Americans on AI use, and maintaining American competitiveness. Notably absent are specific provisions for addressing AI safety or algorithmic bias — areas that some state laws have tackled more aggressively.
The White House explicitly recommends that Congress preempt state AI laws it views as “too burdensome,” arguing that a patchwork of state regulations would hamper innovation. The framework also opposes creating any new federal regulatory body, instead maintaining a “sector-specific” approach through existing agencies.
For privacy advocates, the framework’s call for regulatory sandboxes to accelerate AI development raises concerns about reduced oversight during testing phases. Critics argue the light-touch approach could leave harmful AI applications unchecked until after they’ve caused damage.
Source: CNBC
AI-Driven Layoffs Accelerate: 55,000+ Tech Jobs Cut in 74 Days
The pace of AI-justified job cuts is accelerating. Crypto.com announced it’s cutting 12% of its workforce this week, becoming the latest company to cite AI as a factor in workforce reductions.
The numbers tell a stark story: over 55,000 tech roles have been eliminated across 166 companies in the first 74 days of 2026. AI has been explicitly cited in more than 12,000 of these cuts. McKinsey’s March 2026 Global Institute report found that 12% of job tasks across the economy have been automated by AI over the past two years.
The irony isn’t lost on workers: companies are eliminating jobs to fund AI infrastructure that they claim will eventually create more jobs. Meta’s planned 16,000 cuts and Oracle’s 20,000-30,000 potential layoffs announced earlier this month are funding over $150 billion in combined AI infrastructure spending.
There’s a counterpoint buried in the data: 8% of new job categories created in the same period were directly AI-related. But that’s cold comfort when the jobs being created require different skills than the ones being eliminated.
Open-Source AI Models Close the Gap on Proprietary Giants
Alibaba’s Qwen 3.5, released February 16, is rewriting the competitive landscape for open-source AI. The flagship model ships 397 billion parameters under Apache 2.0 with 256K native context and support for 201 languages — more than double the previous generation’s 82.
This month brought even more open-weight releases: Qwen 3.5 Small arrived March 1 with four dense multimodal models (0.8B to 9B parameters), and Xiaomi’s MiMo-V2-Pro — a trillion-parameter model from former DeepSeek researchers — appeared on OpenRouter March 11 as a free-to-use option.
Meanwhile, DeepSeek V4 remains anticipated rather than available. After missing its mid-February target, reports now point to an April launch. A “V4 Lite” variant briefly appeared on DeepSeek’s website March 9 before being pulled.
For organizations concerned about data privacy, the open-source surge matters. Running capable models locally means sensitive data never leaves your infrastructure. A model that would have been top-5 globally twelve months ago is now available as an open-weight download.
Sources: CNBC, Dataconomy
Quick Hits
-
EU softens AI Act timeline: The Council agreed to extend deadlines for high-risk AI system rules by up to 16 months, waiting until standards and tools are available.
-
Physical AI funding: RoboForce raised $52 million to deploy general-purpose robots across solar, mining, manufacturing, and logistics sectors.
-
AI arbitration hearing: HB 2371, concerning AI-assisted arbitration in divorce proceedings, is scheduled for hearing today (March 23) before the Senate Federalism and Family Law Committee.
-
Alibaba workforce shrink: Alibaba’s headcount continues to decline as the company shifts resources toward its $100 billion AI cloud revenue target over five years.
-
Hunter Alpha mystery solved: The anonymous 1-trillion parameter model that appeared on OpenRouter March 11 was identified as Xiaomi’s MiMo-V2-Pro — now the largest free AI model available.
Worth Watching
The White House AI framework sets up a collision between federal preemption and state-level AI safety laws. States like California and Colorado have passed substantive AI accountability measures that could be overridden if Congress adopts the administration’s recommendations.
The layoff numbers merit ongoing attention. If the current pace continues, 2026 will see more AI-justified cuts than any previous year. Whether AI job creation catches up with job destruction remains the trillion-dollar question.
On the open-source front, the narrowing gap between proprietary and open models could shift enterprise adoption patterns. When a free model matches 89% of GPT-4.5’s performance, procurement decisions get more interesting.