Hacker Attack on OpenAI: Security Vulnerabilities and Internal Tensions

At the beginning of last year, a hacker managed to gain access to the internal messaging systems of OpenAI, the company behind ChatGPT, and steal details about the design of their artificial intelligence technologies. The hacker extracted information from discussions in an online forum where employees talked about the latest technologies from OpenAI. However, the attacker did not infiltrate the systems where OpenAI develops and stores its AI. In April 2023, OpenAI executives informed their employees about the incident during an assembly and briefed the board. However, the public was not informed because no customer data was stolen, and the officials suspected the hacker to be a private individual with no ties to a foreign government. The FBI was not involved either. Some employees then expressed concerns that foreign adversaries, such as China, could utilize the stolen AI technologies, potentially jeopardizing U.S. national security. These concerns raised questions about OpenAI's security management and exposed internal tensions regarding AI risks. After the incident, the technical program manager, Leopold Aschenbrenner, sent a memo to the board, claiming that OpenAI was not doing enough to protect the company from foreign adversaries. Aschenbrenner was later dismissed and argued that his termination was politically motivated. He had hinted about the incident in a podcast, emphasizing that OpenAI's security was insufficient to ensure confidentiality. A spokesperson for OpenAI stressed that Aschenbrenner's security concerns had not led to his separation and that many of his claims did not align with the company's views. Among the emerging AI companies, Meta stands out for openly sharing its technology to find community solutions for potential problems. While today's AI systems can disseminate disinformation and threaten jobs, there is little evidence that they pose a significant threat to national security. Studies by OpenAI and other companies confirm this. Nevertheless, researchers and executives worry that AI could create new bioweapons or breach government systems in the future. OpenAI and other companies have already started securing their technical operations. An OpenAI security committee, which includes Paul Nakasone, the former NSA chief, is exploring how to handle future risks. Government regulations on AI technology are also being planned. Chinese companies are also developing nearly as powerful systems as their U.S. counterparts. Experts warn that these mathematical algorithms, even if they seem harmless today, could become dangerous in the future. "Even if the worst-case scenarios are unlikely, we have a responsibility to take them seriously," said Susan Rice, former advisor to President Biden, at an event in Silicon Valley last month.