What mechanism is most important for ensuring AI safety across international borders, and why?

Most answers emphasize international coordination, independent evaluation, and regulatory oversight as key mechanisms for AI safety, though they differ on whether industry, governments, or new global bodies should lead.

10 answers

Answers

  • Reid Hoffman

    Co-Founder, Executive Board Chair · Manas Ai

    Industry Standards Body

    Reid Hoffman says an industry-run standards body with government oversight is needed, and ideally it becomes international, starting US-led, though he notes international institutions are currently weak.

    Masters of Scale · Watch on YouTube
  • Brad Smith

    President · Microsoft

    Outside Evaluators and Sign‑Off

    Brad Smith says the biggest and maybe most important issue is when outside evaluators should work for a government agency and sign off before a product is released, arguing AI should face safety rules and certification like other products such as planes.

    The answer starts at 6:58.

    The Interview · Listen on Apple Podcasts
  • Elon Musk

    CEO · Tesla

    AI Regulatory Agency

    Elon Musk says there should be some AI regulation, suggesting a standalone AI regulatory agency similar to the FAA or FCC, because the consequences of AI going wrong are severe and we must be proactive rather than reactive.

    Don't Worry About the Vase · Don't Worry About the Vase · Read
  • Jensen Huang

    Founder and CEO · Nvidia

    Third‑Party Evaluators

    Jensen Huang supports third-party evaluators, framed as financial auditors, with the caveat that there have to be multiple.

    Resilient Cyber · Resilient Cyber · Read
  • Dario Amodei

    CEO and Co-Founder · Anthropic

    Embedded Evaluators

    Dario Amodei calls for embedded evaluators from third-party organizations to be given access comparable to internal risk assessment teams, to verify pacing and safety commitments and ensure incidents get reported, and urges governments to require other frontier companies to match.

    Last Week in AI · Last Week in AI · Read
  • David Sacks

    General Partner · Craft Ventures

    Transparency

    David Sacks says transparency and audits are a good idea for regulating AI, but stresses that independent evaluators must truly be objective and neutral rather than cozy arrangements.

    Bloomberg Podcasts · Watch on YouTube
  • Greg Brockman

    President, Chairman, & Co-Founder · Openai

    Third‑Party Auditors

    Greg Brockman does not say third-party auditors are the most important mechanism; he only says the details and nuance matter, that it is too hard to say at a high level whether any particular proposal would work, and notes that third-party auditors already test models before release.

    The answer starts at 19:50.

    Odd Lots · Listen on Apple Podcasts
  • Mark Carney

    24th Prime Minister of Canada | 24e premier ministre du Canada · Office of the Prime Minister of Canada | Cabinet du premier ministre du Canada

    Technology Stability Board

    Mark Carney says a technology stability board modeled on the Financial Stability Board would make sense, arguing coordination is needed because individual companies in a competitive industry find it very difficult to govern safety themselves.

    Bloomberg Podcasts · Watch on YouTube
  • Sam Altman

    Chief Executive Officer · OpenAI

    International Oversight

    Sam Altman says the key is agreeing on rules for development, testing and monitoring alignment standards, with oversight by both countries or an international body to prevent one country from running away with power or breaking the rules.

    The answer starts at 32:01.

    Fortune 500: Titans and Disruptors of Industry · Listen on Apple Podcasts
  • Ian Bremmer

    President and Founder · Eurasia Group

    AI Stability Board

    Ian Bremmer says the key mechanism is an "AI stability board" modeled on the Financial Stability Board, financed by a tax on tokens, to identify, communicate and contain systemic AI risks.

    The Most Interesting Thing in A.I. · The Most Interesting Thing in A.I. · Listen at 17:47