"AI Incidents Will Happen" – Call to Treat Them Like Aviation Incidents

As the development of artificial intelligence systems accelerates, Sam Altman, Chief Executive Officer of OpenAI, stated that preventing all incidents may not be possible, but the way they are handled will determine the safety of this technology.

Speaking at the Dreamforce conference organized by Salesforce in San Francisco, Altman said that “a certain amount of incidents is unavoidable with any new technology,” urging the AI sector to build a safety culture similar to that adopted in aviation.

He explained that airplane accidents are not concealed or treated as isolated failures, but are subjected to thorough investigations to determine their causes and disseminate the lessons learned. Over time, this approach has contributed to making air travel safer.

Similarly, AI companies should transparently report failures in their systems and share information about vulnerabilities and unexpected behaviors, rather than hiding them due to competition or concerns about reputation.

Altman was not advocating for accepting harm, but for acknowledging the possibility of errors and preparing for them before their impacts expand. He also acknowledged that public concerns about losing control over advanced systems or the concentration of power in the hands of a limited number of companies are legitimate.

His remarks come amid growing disagreement within the tech sector between those calling for slowing down the development of more powerful systems and those who believe innovation can continue in parallel with improving safety measures and oversight.