OpenAI’s Latest Technologies Exposed in Breach, Former Employee Alleges Security “Not Strong Enough” Against Foreign Adversaries
OpenAI was hacked in 2023 but kept it quiet. Now, an ex-employee is raising concerns about the breach and OpenAI's security practices. l Source: Win McNamee/Getty Images
Key Takeaways
- OpenAI faced a significant security breach in 2023.
- Hackers accessed internal discussions about their AI advancements.
- Former employees have flagged security concerns and threats on a national level.
Last year, OpenAI experienced a security breach when a hacker infiltrated the company’s internal messaging systems, accessing discussions about OpenAI’s technological advancements.
Although the core systems containing critical artificial intelligence (AI) technologies were not compromised, the incident – and a former employee’s words – raised significant security concerns and potential risks to US national security.
What Happened To OpenAI
Last year, OpenAI faced a security breach when a hacker accessed the company’s internal messaging systems. They obtained details from employee discussions about OpenAI’s technological advancements.
The breach did not compromise systems holding critical AI technologies. However, it raised concerns about security and potential risks to US national security, as reported by the New York Times.
The incident occurred in an online forum where employees shared information on OpenAI’s latest technologies. The breach did not affect core systems storing training data, algorithms, results, or customer information. However, hackers could access some sensitive details.
OpenAI executives disclosed the breach to employees and the board in April 2023. Still, they chose not to publicize it. They cited that no compromise hit customer or partner data, and the hacker appeared to be an individual without government ties. However, this decision faced internal criticism.
Former Employees Resign Over Security Concerns
Leopold Aschenbrenner , a technical program manager at OpenAI, criticized the company’s security measures. He suggested they were inadequate against potential unauthorized access by foreign entities seeking sensitive information.
Following his dismissal for leaking information, which he claimed was due to political reasons, Aschenbrenner publicly voiced concerns about OpenAI’s security practices.
Despite acknowledging his contributions, OpenAI stated that his termination was unrelated to his security concerns.
Aschenbrenner recently discussed the breach on a podcast, marking the first public disclosure of the incident. He raised concerns about OpenAI’s ability to protect critical secrets against potential infiltration by foreign actors, highlighting ongoing debates over cybersecurity measures within the company.
Leopold Aschenbrenner is the latest OpenAI employee to criticize the company publicly. In early June, a letter signed by eleven current and former OpenAI employees and two from Google DeepMind called for better protections for whistleblowers and increased transparency in AI safety.
The letter highlighted the need for improved safety oversight in the AI industry, emphasizing the importance of a “right to warn about artificial intelligence.”
In November 2023, a group of former OpenAI employees accused CEO, Sam Altman, of creating an environment of fear and intimidation, systematically silencing dissent. This letter followed several high-profile resignations. These include co-founder Ilya Sutskever and alignment team leader Jan Leike, who left in May.
Who Hacked Microsoft?
Last month, Microsoft President Brad Smith testified before Congress about Chinese hackers exploiting the company’s systems to launch a broad attack on federal government networks.
However, federal and California laws prohibit OpenAI from discriminating against individuals based on their nationality. Policy experts argue that excluding foreign talent from US projects could severely hinder advancements in AI technology within the United States.
“We rely on the expertise of the most skilled individuals in this field,” said Matt Knight, OpenAI’s head of security. “While this approach carries risks, we must find ways to manage them effectively.”
The Biden administration was preparing to expand its strategy to protect US AI technology from China and Russia. It proposed initial measures to regulate the most advanced AI models, including ChatGPT, as Reuters reported .
