{"id":14364,"date":"2026-09-20T06:01:31","date_gmt":"2026-09-20T06:01:31","guid":{"rendered":"https:\/\/krikis.lt\/openai-saugumas-isklausa\/"},"modified":"2026-09-20T06:01:31","modified_gmt":"2026-09-20T06:01:31","slug":"openai-saugumas-isklausa","status":"publish","type":"post","link":"https:\/\/krikis.lt\/en\/open-security-hearing\/","title":{"rendered":"The OpenAI hack: How security vulnerability disclosure is fueling the AI security debate"},"content":{"rendered":"

Recently, a group of independent cyber-researchers announced that they had hacked OpenAI\u2019s infrastructure. The information, published by Biztoc.com, raises further concerns about the security and reliability of artificial intelligence (AI). While the incident does not reveal any specific data leaks or malicious actions, it does show that even the largest AI companies can have weak spots, especially when it comes to employee accounts and third-party platforms.<\/p>\n

Hacktron study and key findings<\/h2>\n
\"Server<\/figure>\n

According to a source from Biztoc.com, a group of researchers from a small company called Hacktron claim to have hacked OpenAI earlier this year. While the technical details were not disclosed, the researchers say the breach involved Discourse, a community forum platform that OpenAI uses for internal communication. The platform allowed for the transfer of access to employees\u201e ChatGPT accounts, giving them the ability to view sensitive information and even suggest changes to OpenAI\u2019s software.<\/p>\n

PCMag.com provides a more detailed description: Researchers participating in the OpenAI bug bounty program were able to take over the ChatGPT account of an OpenAI employee using a vulnerability in the Discourse system. They emphasize that this access was able to connect to various services such as GitHub, Slack, and email, so the theoretical scope of access was \u201ehuge.\u201d While the researchers say they did not look at sensitive information or upload malicious code, the incident highlights the importance of constantly checking and updating the security of third-party components.<\/p>\n

Fast response and bug bounty reward<\/h2>\n
\"Business<\/figure>\n

Hacktron researchers said in their report that the breach was reported to OpenAI and Discourse within 24 hours, and the issues were fixed within the same period. OpenAI paid Hacktron $$6,500 in bug bounty rewards, according to PCMag.com. This quick response shows that the company takes security vulnerabilities seriously and is willing to invest in fixing them.<\/p>\n

Broader context: AI security initiatives<\/h2>\n

While this particular incident is related to a technical vulnerability, it fits into a broader discourse on AI security. The Times of India reports that Anthropic, along with Accenture, plans to invest $1 billion over five years in independent evaluation of AI models. The initiative comes amid growing pressure from regulators, companies, and researchers to ensure AI models are safe and reliable.<\/p>\n

On the other hand, Nvidia CEO Jensen Huang, as reported by CBS News, expressed the opinion that a coordinated slowdown in AI development is not the right solution, but it shows that major technology companies are actively participating in discussions about AI safety and future risks.<\/p>\n

What does this event mean for the AI community?<\/h2>\n

The Hacktron study reveals several important aspects:<\/p>\n