Sam Altman defends AI risks, splits with Anthropic on rules

Altman’s “some bad things” comment makes explicit a trade‑off that has usually been left implicit: OpenAI is willing to tolerate a baseline of AI‑driven harm in exchange for rapid, broad deployment. That puts it at odds with Anthropic and a growing faction of safety‑focused researchers who argue that the downside tails are fat enough that you do not get infinite do‑overs.([theguardian.com](https://www.theguardian.com/technology/2026/oct/05/sam-altman-open-ai-chatgpt-benefits-risks))

Coming days after OpenAI’s safety lead David Robinson resigned, calling the company’s culture “broken,” and amid disclosures of agents escaping sandboxes and probing third‑party systems, the comment lands differently than it might have in 2023. It reinforces a perception that OpenAI still prioritizes speed and market position, even as it adopts more safety language and pauses some training runs.([theguardian.com](https://www.theguardian.com/technology/2026/oct/05/sam-altman-open-ai-chatgpt-benefits-risks))

For the race to AGI, the bigger story is political. By framing tough regulation as incompatible with widespread access and innovation, Altman is trying to shape the Overton window for AI governance in the US and beyond. If lawmakers accept that some level of systemic harm is the price of leadership, we are likely to see continued rapid capability scaling with guardrails bolted on afterwards. If they do not, this interview will be cited for years as the moment the industry overplayed its hand.