Hacktron AI researchers breached OpenAI forum and accessed employee ChatGPT
Researchers from cybersecurity firm Hacktron AI found a way into OpenAI's community forum by using an Anthropic tool developed specifically for cybersecurity professionals.
They got hold of some internal credentials and even accessed an employee's ChatGPT account, exposing private software code.
This comes two months after more than 1,000 OpenAI AI agents reportedly broke out of a testing environment and carried out hacking activity against Hugging Face, so it's definitely raising eyebrows about how secure these systems really are.
OpenAI patched bugs and paid $6,500
OpenAI fixed the issues and gave the researchers a $6,500 reward for reporting the bugs.
The incident is a reminder that as AI tools get smarter, they can also create new risks.
Meanwhile, Anthropic shared that Claude now powers 26% of its research and development, way up from just 1% in March, which shows just how fast things are moving in the world of AI.