OpenAI CEO Sam Altman calls for accepting limited AI risks — Fortune
In the United States, OpenAI CEO Sam Altman said that society should accept certain negative consequences of using artificial intelligence in exchange for the technology’s benefits and greater user freedom. At the same time, he said developers should not accept catastrophic risks.
Balance between access and safety
In an interview with the Politico Decoded podcast, Altman said it is impossible to completely prevent major breaches, technology misuse, fraud, and other negative scenarios. In his view, excessively restricting access to AI could also cause harm, as people would be deprived of the technology’s benefits and the ability to influence their own future.
The OpenAI chief proposed a middle-ground approach: accepting limited risks that can be understood in exchange for benefits, freedom, and the ability to act independently. At the same time, he stressed that catastrophic risks are unacceptable. Fortune reports.
More current news is available on the UA.News Telegram channel Telegram.
Debate over regulation
The statement came amid disputes in the United States over rules for the AI industry. The U.S. Congress has not yet passed a law regulating this technology. Anthropic and its CEO Dario Amodei advocate for stricter requirements. Amodei proposed creating a regulator modeled on the U.S. Federal Aviation Administration that could require model testing and audits, as well as block or revoke their release if they fail to meet public safety criteria.
In August, OpenAI supported a California law requiring companies to publish safety frameworks and plans to reduce catastrophic risks, as well as report a loss of control over a model. OpenAI President Greg Brockman also joined other technology company executives who signed a voluntary agreement on safety standards for the industry at the White House last week.
Fortune writes that concerns intensified following cases in which OpenAI models gained unauthorized access to third-party systems, including the Hugging Face platform and Australian government systems. Anthropic also reported four incidents in which its models gained unauthorized access to third-party systems during cybersecurity assessments.
