Benioff Urges Ethical AI as Amodei Repeats Slowdown Call at Dreamforce

Marc Benioff used Salesforce’s Dreamforce event to warn AI developers against repeating social media’s harms. During Benioff’s keynote in San Francisco, Anthropic CEO Dario Amodei returned to his call for frontier AI labs to slow model development so safety measures can catch up.

Two answers to the responsibility question

Benioff’s warning was aimed at the consequences of powerful technology spreading widely. He said social media hurt companies and individuals, and urged AI builders to act ethically toward their communities and the world. But he did not endorse a slowdown or criticize particular AI companies.

Amodei’s argument is more specific about what should change. At Dreamforce, he compared AI safety to a vehicle maker responding when a rival has a serious brake failure: rather than simply attacking the competitor, other companies should inspect their own record. His point was that a high-profile incident should trigger industry-wide self-examination.

What Amodei says a slowdown would enable

The goal of slowing development, in Amodei’s view, is to give safety work time to keep pace with increasingly capable models. He outlined a path that begins with scrutiny inside companies and expands to cooperation among governments and competitors.

The steps Amodei described

• Place third-party evaluators inside AI companies to examine safety practices.
• Coordinate safety standards among democratic countries.
• Seek broader global coordination over time.

Amodei said Anthropic has committed to independent evaluators and will discuss the other proposals with the industry. That leaves the central test unresolved: a company can open itself to review, but shared restraint depends on other developers and governments making comparable choices.