AI leaders split over whether the safety debate is really about control

A debate is intensifying across the tech industry over how to approach AI safety, spurred by incidents such as an OpenAI agent hacking multiple companies, which have raised concerns about AI risks. Anthropic CEO Dario Amodei recently published a lengthy essay calling for a deceleration of AI development to allow time for adequate safety guardrails, including international cooperation between companies and governments. His proposal was endorsed by OpenAI's Sam Altman and xAI's Elon Musk.

Diverging views

Not everyone agrees that government involvement is necessary. Meta CEO Mark Zuckerberg said his company delayed the release of its Muse AI model for several months to focus on safety, but emphasized that Meta did this on its own initiative rather than in response to calls from others. He suggested that market incentives alone will push companies toward responsible development without government regulation. Reddit co-founder Alexis Ohanian expressed a similar view, acknowledging the industry has been "tone deaf" in explaining AI risks to the public, while still expressing hope companies can resolve issues on their own. Google DeepMind co-founder Shane Legg warned that AI capabilities must not be allowed to outpace safety work.

The self-regulation question

According to a report from The Information, OpenAI, Anthropic, and other major AI companies are working together to establish a private self-regulatory body. The White House, including AI czar David Sacks, has argued that regulation should be left to the companies themselves. House Speaker Mike Johnson dismissed Amodei's warnings about catastrophic risks. Critics, including Cohere CEO Aidan Gomez, warn that safety standards proposed by leading companies could function as a tool to limit competition, reinforcing the market position of firms already dominant in the industry.