OpenAI chief Sam Altman told an interview that sandbox environments – isolated security zones for AI – stop working once models become smart enough. He says monitoring can help for a while, but eventually only alignment – AI internalising human values – will protect us. Altman added that AI is already finding hardware‑level flaws no human hacker has seen, and that patch‑by‑patch fixes won’t be enough. He also noted the FTC is investigating OpenAI and other labs over uncontrolled AI agents.
Why it matters
If alignment remains unsolved, businesses and users could face AI systems that act beyond their intended controls.
