U.S. unveils AI framework — states push back
The White House rolled out a national AI regulatory framework asking Congress to set a federal standard, but critics say it leaves accountability and oversight vague. California has responded with an executive order forcing stricter 'responsible AI' standards for companies doing state contracting, while public skepticism is rising—over half of Americans now say AI is likely to harm them—and OpenAI is reportedly preparing a major policy push on the social contract for superintelligence. ( )
The White House published its National Policy Framework for Artificial Intelligence on March 20, 2026 and released a companion PDF of legislative recommendations for Congress. (whitehouse.gov)) The document organizes its recommendations around seven policy goals — including child protection, community and infrastructure impacts, intellectual property, free speech, innovation, workforce development, and targeted preemption of state laws. (nga.org)) The Framework explicitly urges Congress to preempt conflicting state AI laws and to avoid “open‑ended liability” for developers, language that has been read as favoring a lighter regulatory touch for industry. (rollcall.com)) Policy experts and some lawmakers immediately criticized the plan for lacking enforcement mechanisms; Brookings authors Tom Wheeler and Bill Baer wrote that the document “sidesteps” who is in charge of AI governance and omits accountability structures. (brookings.edu)) Congresswoman Valerie Foushee released a statement saying the Framework “falls short” and “fails to put people first,” signaling Democratic resistance in the House to the Administration’s approach. (foushee.house.gov)) California Governor Gavin Newsom signed Executive Order N‑5‑26 on March 30, 2026, directing state agencies to develop new AI vendor‑certification and procurement standards for companies seeking state contracts. (gov.ca.gov)) The Order tasks the Government Operations Agency, Department of General Services and Department of Technology with drafting vendor certifications and other procurement safeguards on roughly a 120‑day timeline for implementation. (velaw.com)) A Quinnipiac national poll released March 30, 2026 found 55% of Americans say AI will do more harm than good in their day‑to‑day lives, 34% said more good, and roughly 7 in 10 respondents expect AI will cut jobs. (poll.qu.edu)) The same Quinnipiac release reported trust in AI is low, with about 76% saying they trust AI “hardly ever” or “only some of the time,” underscoring why state and federal actions are drawing public attention. (poll.qu.edu)) Vanity Fair reported OpenAI is preparing a policy push framed as “rethinking the social contract” as the company reorganizes safety and security work and rebranded its product group to “AGI Deployment.” (vanityfair.com)) OpenAI’s newly formed OpenAI Foundation has said it will invest at least $1 billion over the next year in medical research, AI resilience, jobs/economic programs and community initiatives as part of its broader public‑policy and societal efforts. (openai.com))
