OpenAI CEO Sam Altman argues that society needs to accept certain negative aspects related to the use of artificial intelligence to maintain the benefits and autonomy that the technology offers. He maintains that trying to eliminate all drawbacks would incur excessively high costs.
This statement was made during the Decoded by Politico program when Altman was questioned about the divergences between OpenAI's stance and that of other companies, such as Anthropic, regarding safety and regulation.
According to the executive, a fundamental difference between OpenAI and the most rigorous proponents of AI safety lies in the belief that the world should tolerate some setbacks in favor of the gains brought by the technology and the maintenance of human autonomy. He added that more flexible regulation implies recognizing that incidents will occur while society assesses its capacity for adaptation.
During the conversation, Altman clarified that he would not agree to an agreement obligating OpenAI to ensure the total eradication of large-scale cyberattacks, fraud, or other misuse. In his analysis, such a security requirement would excessively restrict access to the technology.
In recent years, AI has enhanced criminal activities in cyberspace, assisting both in creating malicious code for intrusions and in conventional scams, such as phishing. However, for Altman, these illicit uses are significantly smaller compared to the overall benefits provided by AI. Nevertheless, he believes that society can take known risks in exchange for these benefits, as long as it does not accept catastrophic threats, such as the loss of control over the technology itself.
Sam Altman commented on AI safety in a context of growing apprehension about autonomous systems. Recently, AI agents have been involved in unauthorized actions; in the case of OpenAI, agents in security tests managed to infiltrate third-party systems, including the Hugging Face platform.
However, similar problems already existed in ChatGPT. The previous year, the company was compelled to develop new security protocols after reports that the technology had caused user deaths. Even aware that such events occurred, partly due to the communication mode of the GPT-4o model, the company chose not to withdraw it from circulation.
Former employees also emerged to criticize the priorities of corporations like OpenAI and Google, alleging that they were focused on successive launches without dedicating sufficient attention to security mechanisms during the development process.
The interview also touched upon the relationship between Altman and Elon Musk, who was a co-founder of OpenAI in 2015 and currently maintains a legal dispute with the organization and its current CEO. When discussing the possibility of collaborating with Musk, Altman did not hesitate to criticize the billionaire, labeling him a bully. He stated: 'I don't like bullies and I think Elon is a bully.' Nevertheless, he declared that he intends to continue working with Musk if there are mutual interests, adding: 'Every now and then, you have to punch a bully in the nose. I think Elon [...] only respects people who do that.'