Amodei urges AI slowdown; Altman and Musk in, Sacks dissents
TL;DR
- Anthropic CEO Dario Amodei called for slowing frontier AI development in a September 12 blog post; OpenAI's Sam Altman and Elon Musk agreed.
- Amodei proposed embedded third-party evaluators, safety-standard coordination among democracies, and compliance verification with authoritarian governments.
- Trump AI official David Sacks backed the decision but called the altruism framing cover for product-liability exposure.
Anthropic CEO Dario Amodei has asked the industry to slow itself down. In a September 12 blog post titled 'We Must Pace the Frontier,' he wrote that companies 'must slow the pace at which we improve the capabilities of AI models,' clarifying that 'pacing does not mean halting model training or technical progress, but ensuring companies take adequate time to align and safeguard their models, and for third party evaluators to confirm this.' OpenAI's Sam Altman and Elon Musk publicly agreed.
David Sacks, a Trump administration AI official, did not.
Writing at Tech Policy Press, Justin Hendrix opens on 'the frenzy of headlines about AI safety over the past several days,' a stretch that includes Anthropic disclosures involving Iranian attempts to target US Navy warships and Houthi work on ballistic missiles using its models. Amodei's answer arrives as three proposals: ongoing access for embedded third-party evaluators, coordination among democratic countries on common safety standards, and an effort to bring authoritarian governments into compliance verification.
Sacks read the pitch differently. 'If the unreleased models are scary enough that you think you should slow down, I support your decision,' he wrote, but told the labs to 'stop pretending you need anyone else's permission' and to 'stop pretending the motivation to slow down is purely altruistic.' His reason: 'You face massive product-liability exposure if your products enable a truly damaging cyberattack.'
The other set of actors named in the piece is the one not moving. Hendrix reports that President Trump remains unconcerned with AI risks and focused on competing with China, that House Speaker Mike Johnson has put the duty of AI safety on companies rather than Congress, and that former President Barack Obama has urged Democrats to make AI a central issue if they take back the House. Legislation before the midterms looks unlikely.
The headline's question, who actually sets the pace, Hendrix leaves standing.
Shared on Bluesky by 2 AI experts
-
A slew of Anthropic misuse stories, another OpenAI agent swarm hack, and CEOs calling for a slowdown — all in a few days. In this week's Tech Policy Press newsletter, Justin Hendrix considered a news cycle moving faster …
View on Bluesky →
Originally reported by techpolicy.press
Read the original article →Original headline: 'Pace the Frontier,' They Say. But Who Sets the Pace?