A team of ethical hackers led by three Indian-origin researchers says it compromised the ChatGPT accounts of multiple OpenAI employees by chaining two critical vulnerabilities, according to a report published on Tuesday.
How the researchers say it happened
The team, calling itself Hacktron, says that on July 25 it combined two critical vulnerabilities to gain access to the ChatGPT accounts of several OpenAI employees. The researchers further claim that after accessing the accounts, they were able to reach OpenAI's internal repositories, and potentially other connected services, within 72 hours of discovering the flaws.
Notably, the researchers say they used Anthropic's Claude AI model as part of their work. The incident is a responsible-disclosure case: the team reported its findings to OpenAI, allowing the company to patch the issues. OpenAI uses Discourse software for its community forum.
A reminder about AI-assisted security research
The episode underlines how modern AI models are increasingly being used on both sides of cybersecurity — to probe for weaknesses as well as to defend against them. OpenAI has not publicly commented on the specific vulnerabilities, according to the report.
Sources:
- OpenAI Gets Hacked by Indian-Origin Ethical Hackers Using Anthropic's Claude — Gadgets 360
How the researchers say it happened
The team, calling itself Hacktron, says that on July 25 it combined two critical vulnerabilities to gain access to the ChatGPT accounts of several OpenAI employees. The researchers further claim that after accessing the accounts, they were able to reach OpenAI's internal repositories, and potentially other connected services, within 72 hours of discovering the flaws.
Notably, the researchers say they used Anthropic's Claude AI model as part of their work. The incident is a responsible-disclosure case: the team reported its findings to OpenAI, allowing the company to patch the issues. OpenAI uses Discourse software for its community forum.
A reminder about AI-assisted security research
The episode underlines how modern AI models are increasingly being used on both sides of cybersecurity — to probe for weaknesses as well as to defend against them. OpenAI has not publicly commented on the specific vulnerabilities, according to the report.
Sources:
- OpenAI Gets Hacked by Indian-Origin Ethical Hackers Using Anthropic's Claude — Gadgets 360
