AI Regulation & Policy

Today's Stories

1. 1

   Sanders, Casar bill would ban artificial superintelligence

   Sen. Bernie Sanders and Rep. Greg Casar unveiled a bill Wednesday that would permanently ban artificial superintelligence, temporarily pause the most advanced AI development until federal safety rules are set, and create a Department of Artificial Intelligence. Violations could carry up to 20 years in prison in some cases. Multiple employees at leading AI companies are endorsing the bill, giving the push for AI guardrails support from inside the industry even as President Donald Trump has resisted oversight and argued mainly about the economic cost of a slowdown.

   The bill faces long odds in the Republican-controlled Congress and the prospect of slowing US development while China continues has been a central argument against sweeping restrictions. AI safety is expected to be a central part of talks between Trump and China's Xi Jinping this week.

2. 2

   Trump slams AI oversight at UN as Sam Altman heads to Security Council

   Trump told the UN General Assembly AI oversight is a 'globalist scheme,' while Sam Altman and Dario Amodei are set to brief the UN Security Council on AI and international security a day later. The two speakers are set to ask the UN for close to the opposite of what Trump said, putting the labs' leaders at odds with his stance.

   The outcome hinges on whether the Council session shifts any national position, and on Trump's meeting with Xi Jinping, where AI is expected to be discussed.

3. 3

   Pope's AI Adviser Warns of 'Cartel' Rule-Setting by Big Labs

   Father Paolo Benanti told WIRED that warnings about superintelligence from labs like Anthropic and OpenAI distract from public debate on controlling the companies, calling their self-interested rule-setting a 'cartel.' Benanti argues ethics and standards should not be owned by company boards, and wants democratic regulation and mandatory third-party evaluations instead of bans on techniques such as distillation.

   Anthropic declined to comment, while OpenAI and Google DeepMind did not respond; Pope Leo XIV's first encyclical already urged AI companies and rival superpowers to set aside geopolitical or commercial dominance.

4. 4

   Sanders, Casar bill: 20 years prison for superintelligence

   Sen. Bernie Sanders (I-VT) and Rep. Greg Casar (D-TX) introduced the Ban Artificial Superintelligence Act, which would ban developing artificial superintelligence and impose up to 20 years in prison on violators. The bill treats the most advanced AI as a threat serious enough to jail the people building it, a penalty the bill compares to unlawfully developing nuclear weapons.

   The proposal is likely to face political headwinds, as potential regulation still seems shaky overall despite recent moves like California Governor Gavin Newsom's proposed "kill switch" for frontier models.

5. 5

   Newsom signs seven bills to make data centers disclose energy, water use

   California Gov. Gavin Newsom signed seven bills, including AB 1577 requiring monthly energy reporting and AB 2619/AB 2469 requiring water disclosures, starting next year. The transparency is meant to show whether data centers are raising electricity bills, draining water, and meeting sustainability commitments—questions researchers say they couldn't answer without basic facts.

   Water disclosures only happen at permit or license application, not yearly, so the data may reflect a wet or cool year rather than actual annual use; implementation details are still being worked out.

6. 6

   Gallup: 68% of daily AI users in US worry

   A Gallup survey commissioned by Microsoft, polling about 1,000 people in each of 37 countries from April to July, found 68% of Americans who use AI daily are worried about it, and 74% of Americans overall reported worry. Even daily use does not imply comfort, so adoption is not a straightforward measure of public enthusiasm and may not reduce support for regulation.

   The study will eventually span 140 countries and currently omits India, Australia and Malaysia, so the picture could shift; watch the 36% (median) who said they trust AI's results a lot as a sign of whether confidence catches up.