Security researchers from Hacktron AI successfully breached OpenAI systems by exploiting vulnerabilities in the company's community forum. This access allowed them to compromise an employee's ChatGPT account, which contained credentials for internal code repositories on GitHub.
The team utilized specialized AI tools developed by Anthropic to identify these security flaws. OpenAI confirmed the breach and stated that the vulnerabilities have since been patched. The researchers received a bounty of $6,500 for their findings as part of a bug bounty program.
Meanwhile, Anthropic reported a significant increase in the use of its Claude AI for internal research and development. The model now assists in 26 percent of projects, compared to just 1 percent in March. The company aims to highlight the progress toward recursive self-improvement.
Anthropic noted that while AI is taking on more complex tasks, humans remain involved in 90 percent of operations. Despite concerns regarding the potential loss of control over autonomous systems, the company maintains that current AI models still require human oversight.
