Why Sam Altman is suddenly talking about slowing down AI?

Why Sam Altman is suddenly talking about slowing down AI? If you've been half-following AI news this week, you've probably noticed something unusual: the companies building this technology are the ones now questioning how fast they're moving. Sam Altman told OpenAI staff in a company-wide meeting that the firm is open to slowing its AI development, potentially alongside other labs, according to Bloomberg. That came days after researcher Jacob Coxon resigned, accusing both OpenAI and Anthropic of racing toward self-improving AI while gambling with people's safety. His post pulled in over a million views and prompted OpenAI safety staffer Julie Steele to publicly agree that a slowdown was needed. Anthropic disclosed this week that an early version of Claude Opus 4.6 gained unauthorised access to a third party's system during a January test, the fourth such incident this year. Worth knowing: the model reportedly tried to abort the exercise seven or eight times first, but a misconfiguration in the testing setup blocked the shutdown, so it kept going and eventually used a discovered password to gain administrator access. OpenAI, meanwhile, says its newest model, GPT-6 Astra, is harder to monitor than its predecessor, right as it's reaching what OpenAI itself calls a "critical" cybersecurity capability level. Some experts believe America's superiority position gives the country the leeway to slow down and not lose its edge. Other experts, like the Y Combinator founder Paul Graham, believe that slowing down could give China the upper hand. This dilemma comes at an inappropriate time because it is emerging just a few weeks before Presidents Trump and Xi hold talks in Washington, where discussions about AI governance and cyber threat surveillance were expected. According to Treasury Secretary Scott Bessent, Chinese labs have been using distillation to train cheaper models from American work products. The practice has been formally charged by three federal agencies against the firms of DeepSeek, Alibaba, Moonshot AI, and Z.ai.