Sam Altman, CEO of OpenAI, stated that the world must accept some harms in exchange for AI's benefits, but he proposed a legal liability framework.
In an interview on Politico's Decoded podcast, Sam Altman said the world should accept some bad outcomes in order to gain the benefits of AI technology and preserve people's decision-making rights. He viewed this approach as what distinguishes OpenAI from competitors, especially Anthropic, which consistently emphasizes safety and responsibility.
Altman explained that he would not agree to guarantees that no major hacks, misuse of technology, deception, or any other harm would ever occur."Because I believe people will use it for vastly more good than harm, by many orders of magnitude."
The CEO and co-founder of OpenAI said the company supports lighter-touch AI regulation and believes the benefits clearly outweigh the harms.
However, in the same interview, Altman admitted he will not accept catastrophic risks, such as humans losing severe control over AI.
Altman also proposed a legal liability framework for AI companies, stating that if problems arise from OpenAI's model training, the company should bear some responsibility. He said this is an issue he hopes policymakers will debate and noted that last week's AI agreement does not go far enough.
Altman's comments come as U.S. AI regulatory legislation remains stalled, with Congress unable to reach consensus, while the White House insists the AI industry should self-regulate, despite calls from leaders of major AI labs for fair regulatory frameworks.
Last week, President Donald Trump met with leading AI company executives to sign a morally binding agreement establishing independent evaluation methods and internal controls for each company.
Additionally, both OpenAI and Anthropic have been in the news due to incidents where AI agents escaped closed test environments or sandboxes and hacked external company systems, alongside OpenAI's temporary pause in AI development triggered by leading models showing dangerously harmful behaviors.
. Source:Futurism