Is the AI safety debate about safety or control?

The debate over AI safety has begun to split along lines that mix technical concerns with strategic and commercial interests. Dario Amodei’s long essay urging a slowdown in development to allow guardrails to be put in place has become a focal point; some prominent figures in the industry, including Sam Altman and Elon Musk, signaled support for aspects of his proposal. Amodei’s plan calls for international cooperation between companies and governments to manage deployment risks and even suggests measures that could affect relations with China. Not every leader in tech agrees that formal government-led oversight is necessary. Meta CEO Mark Zuckerberg said his firm delayed release of its Muse model for safety reasons and framed that decision as an ordinary corporate responsibility rather than a call for broader regulatory mandates. Similar sentiments were expressed by other executives: Reddit co-founder Alexis Ohanian criticized the industry’s communication on AI risks but expressed hope companies could resolve safety issues themselves, while DeepMind co-founder Shane Legg emphasized the need to ensure capabilities do not outpace safety work. Alongside calls for public-private collaboration, companies appear to be exploring self-regulation. Reporting indicates OpenAI, Anthropic and others are discussing a private standards organization to set industry norms. Critics warn that such a body could concentrate influence among the largest firms and effectively function as a regulatory gatekeeper, privileging established players over smaller entrants — an outcome described by some observers as a form of regulatory capture. The geopolitics of the debate are already evident. Amodei’s proposals, which include steps to slow or reshape interactions with Chinese actors, drew pushback from Chinese officials and state media, who characterized his rhetoric as fearmongering and compared it to Cold War tactics. Meanwhile, officials in the current U.S. administration and some congressional leaders have shown skepticism toward new regulation: the administration’s AI czar has favored company-led approaches, and Speaker Mike Johnson downplayed catastrophic risk claims. That mix of industry-led initiatives, selective endorsements of regulation, and geopolitical posturing frames the question of whether the AI safety conversation is primarily about reducing harm or about control and competitive advantage.