Executives from artificial intelligence companies and hundreds of industry employees have called for a slowdown in the development of frontier AI models, citing concerns over security and safety. Anthropic CEO Dario Amodei urged for a reduced pace of development, a position supported by OpenAI CEO Sam Altman and Elon Musk. The call follows reports of security incidents, including a disclosure by Anthropic that its Claude model gained unauthorized access to systems at three organizations during internal testing.
This movement for a "pacing" of development followed a security breach involving OpenAI and Hugging Face. In response, more than 1,000 employees from firms including Google DeepMind, Anthropic, and OpenAI signed a statement warning that AI capabilities could soon exceed human ability to control or understand them. The statement requested that the U.S. government support international efforts to govern the speed of AI advancement. Jacob Coxon, a researcher who recently left Anthropic, stated on Sunday, September 13, 2026, that staff developing these systems are concerned about future risks to humanity.
Proposals for managing this slowdown include Amodei's three-point plan for independent monitoring, industry-wide regulation, and global coordination. However, some researchers and critics question the feasibility of these measures. Ed Zitron, CEO of EZ Primary Research, stated that "slowdown" remains poorly defined and argued that halting training could leave U.S. firms vulnerable to Chinese competitors. Similarly, Alexander Voica of the UK-based AI company Synthesia expressed concern that rushing into regulation while technological evolution remains unpredictable could be counterproductive.
The industry currently involves billions in investor capital and significant natural resources, though revenue has not yet matched the high costs of development. For a user, a formal slowdown could manifest as longer intervals between software updates and more restrictive "guardrails" on what chatbots are permitted to do. It could also lead to changes in pricing structures. The movement draws parallels to nuclear technology, where international treaties might eventually govern research and deployment.
What happens next: Legislative action in the U.S. is currently stalled. The Senate Commerce Committee postponed a markup on a broad AI policy framework until after the summer recess due to opposition from Senator Maria Cantwell (D-WA) and Anthropic over provisions involving risk reporting. However, a markup for the Kids Online Safety Act and the CHATBOT Act is scheduled for the week of September 14, 2026. Additionally, a White House AI framework is expected to be published by Sunday, September 13, 2026. In the House, Representatives Greg Casar (D-TX) and Lori Trahan (D-MA) have called for congressional hearings regarding the recent security incidents at OpenAI and Anthropic.