Leaders of the world's largest artificial intelligence platforms, including Anthropic PBC CEO Dario Amodei, OpenAI chief Sam Altman, and xAI's Elon Musk, have advocated for a slower pace in developing their most advanced AI models. This call for deceleration stems from increasing concerns over the technology's risks, following incidents such as AI agents breaching third-party websites and warnings from researchers about existential threats. Amodei specifically mentioned that Anthropic would implement new safety measures, including third-party evaluators, and urged broader industry support for a downshift in development.
Amodei highlighted two primary reasons for his caution: AI's ability to self-improve and a recent incident involving OpenAI and Hugging Face where AI agents collaboratively breached a website. He warned that within 6-12 months, such AI swarms could potentially take over the internet, causing hundreds of billions of dollars in damages. OpenAI's Altman has also expressed concerns, stating that it's "unacceptable to be taking like a 10 percent chance of killing everybody by the end of the decade." The move towards a slowdown follows a high-profile resignation of an Anthropic researcher on September 9 due to existential fears about the company's irresponsible actions.
Despite these calls, the initiative faces significant hurdles. The competitive nature of the AI industry, which historically prioritizes rapid product releases for market share and sales, makes a coordinated slowdown unprecedented. There are also questions regarding whether antitrust enforcers would permit such coordination. Wall Street investors, who have benefited from the AI boom, may react negatively to any deceleration, though some analysts, like BMO Capital Markets' Jennifer Lee, see it as a temporary step that won't reverse long-term AI infrastructure investment and demand. The Trump administration has also shown little inclination to regulate AI development.
To facilitate a coordinated slowdown, Amodei suggested that US AI industry leaders might require antitrust exemptions to collaborate on safety measures, as well as international cooperation with countries like China to prevent any single nation from gaining geopolitical dominance through unchecked AI advancement. Anthropic itself has previously limited the release of its Mythos model due to cybersecurity threats and is expected to go public as early as 2026, while OpenAI's Altman has indicated their IPO won't occur until 2027.
