Is Big Tech’s AI slowdown a safety pact or a cartel?

Your AI coding assistant's capabilities might be intentionally limited to prevent existential risks, or it could be a tactic to slow competitors.

Leaders from major AI companies, including OpenAI, Anthropic, and Google DeepMind, have publicly discussed a potential slowdown in AI development. This initiative, partly inspired by proposals from Anthropic CEO Dario Amodei, suggests measures like third-party auditing and international agreements to "pace the frontier." Critics, however, question these motives, suggesting it could be an attempt to stifle competition and open-source AI, likening it to a "cartel." Concerns about AI safety have intensified following reports of internal dissent and security breaches at leading labs. While some industry observers see the discussion as a positive step, others worry it might be "safety-washing"—a superficial effort to appear responsible without substantial change. The prospect of AI reaching recursive self-improvement (RSI), where models can advance independently, fuels these urgent discussions. The geopolitical competition with China is frequently cited as a counterargument to any slowdown, with fears that the US falling behind technologically.

The public declaration by AI leaders to "pace the frontier" presents a complex signal, simultaneously addressing genuine safety concerns raised by researchers and potentially serving as a strategic move to consolidate market dominance. The voluntary nature of this agreement, without immediate regulatory backing, raises questions about its enforceability and whether it will lead to meaningful change or merely a performative gesture. The industry's history of lobbying for self-regulation to preempt less favorable external laws suggests a pattern of prioritizing competitive advantage under the guise of safety. This dynamic positions AI development at a critical juncture, where the motivations behind slowdown proposals need careful scrutiny to distinguish genuine risk mitigation from market control strategies.

The calls for a slowdown, particularly from established frontier AI labs, could inadvertently create an environment that favors large, well-funded organizations over smaller, agile competitors and the open-source community. This could manifest as a tacit agreement to avoid rapid advancements that might trigger stringent, potentially stifling, regulations, thereby cementing the current leaders' positions. The stated need for "new champions" in AI safety discussions highlights a potential disconnect between the industry's self-perception and the broader ecosystem's desire for independent oversight. Moving forward, watch for whether these discussions translate into concrete, verifiable actions that slow development across the board, or if they primarily serve to create barriers to entry for emerging AI ventures.