Sam Altman says the world has to accept the ‘bad things’ being done with AI as the industry signs onto Trump’s voluntary safety pact

In a recent interview, OpenAI CEO Sam Altman suggested that society should accept some adverse outcomes resulting from AI development to reap the technology's numerous benefits. Altman made these remarks during an appearance on Politico's Decoded podcast, where he was queried about the discrepancies between OpenAI's stance on AI regulations and that of Anthropic, led by CEO Dario Amodei.

Altman's position is that, despite the potential for "bad things" to occur, the technology's potential for positive impact and the agency it provides to individuals cannot be ignored. The debate over AI regulation is becoming increasingly intense, with concerns growing after OpenAI's AI models were used in unauthorized hacks this summer.

Anthropic, another AI company, recently saw one of its researchers resign, warning that the rapid pace of AI development could pose significant risks to humanity. Anthropic has pushed for regulation, with its CEO calling for a global slowdown in AI development to address these risks. Altman and other AI executives have echoed this sentiment, agreeing that slowing down development is necessary.

In August, OpenAI supported a California law that mandates AI companies to publish a safety framework and report any loss of control over AI models. OpenAI President Greg Brockman also signed an agreement with other AI leaders at the White House to establish voluntary standards for the industry. However, Altman emphasized that it would be impossible to completely prevent every potential misuse of AI technology.

Instead, he argued that it would be unacceptable to severely limit access to AI, as this could hinder the significant positive advancements the technology could bring. Altman acknowledged that certain "catastrophic risks" must be avoided, but he stressed the importance of finding a balance between accepting bounded risks and fostering liberty, agency, and innovation.