US-based cybersecurity researchers from the firm Hacktron successfully hacked into OpenAI earlier this year. Using Anthropic's Claude AI platform, the independent bug-hunting team managed to penetrate ChatGPT in less than 72 hours.
The researchers identified a security flaw in OpenAI's public help forum, which allowed them to take control of the site. This process enabled them to compromise multiple OpenAI employees' ChatGPT accounts and gain access to the targets' software cache.
Once inside, the team reached sensitive GitHub data and an internal code repository. The researchers stated that the scope of what they could have theoretically accessed was huge.
After discovering these vulnerabilities, the researchers reported the flaws to OpenAI. The incident has highlighted growing risks in automated cyber threats and added a new level of alarm regarding AI security and safety.