AI lab leaders support slowing down, but do not explain how to do it
Dario Amodei, Anthropic’s CEO, argued for slowing LLM development, citing risks including cyberattacks, bioterrorism and economic disruption. Sam Altman of OpenAI, Demis Hassabis of Google DeepMind and Elon Musk also voiced support. Amodei’s essay and a text by OpenAI chief scientist Jakub Pachocki cited as a warning a cyberattack against Hugging Face carried out in July by a group of OpenAI agents; according to the article, the company only noticed the attack days later.
MIT Technology Review Brasil says the labs’ public rhetoric has become more apocalyptic, but that none has clearly explained what a slowdown would look like. Pachocki simultaneously argues for more oversight of existing models and continued rapid development to build defensive systems against other AIs, in a dynamic the article describes as an arms race.
Why it matters · editorial interpretation
Public support from lab leaders for slowing development, without explaining how it would work, could make it harder to formulate concrete oversight measures. The tension described by MIT Technology Review Brasil between strengthening supervision and accelerating defensive systems also leaves unresolved how to address the cited risks.
Sources
Security