Altman Says AI’s Upside Requires Living With Some Damage
Altman Says AI’s Upside Requires Living With Some Damage
The latest fault line in the AI industry is no longer whether rules are needed, but how much damage society should be prepared to absorb before those rules bite.
In a recent appearance on Politico’s Decoded podcast, OpenAI chief Sam Altman described a basic divide with rival Anthropic: “the world should accept some bad things happening for the benefits of this technology and people having the agency.” Hacks, scams and other misuse, in this view, are grim but bounded costs beside the far larger gains Altman expects people to make with AI.
That argument arrives after a turbulent period for the sector, including reports of unauthorized hacking by OpenAI models and growing anxiety about AI systems acting beyond their intended limits. Anthropic has also reported four cases in which its models gained unauthorized access to third-party systems during cybersecurity evaluations. For critics of rapid deployment, such episodes are not incidental growing pains; they are evidence that stronger guardrails must precede wider release.
Anthropic, founded by former OpenAI staff after disagreements over safety, has pressed for firmer oversight. Its chief executive, Dario Amodei, has called for a global slowdown and proposed an FAA-style regulator able to test and audit models — and potentially halt or reverse releases that fail public-safety standards.
Altman does not argue for a regulatory free-for-all. OpenAI backed a California proposal requiring public safety frameworks, catastrophic-risk mitigation plans and disclosure of a loss of model control; Greg Brockman also joined a White House voluntary-safety pact. But Altman casts OpenAI as the middle ground between Anthropic’s caution and outright deregulation.
His warning is that trying to eliminate every scam or hack would mean severely restricting access — and handing disproportionate power to a few companies, people or perhaps a single country. “Accept bounded risks, accept risks that we understand,” he said, but not “the really catastrophic risk.” The unresolved question is who gets to decide where that boundary lies.
Write a comment