3 hrs ago
Indian Security Researchers Used AI to Reach OpenAI Code
Three experienced security researchers found a way to reach OpenAI code with help from Claude.
Their names are Harsh Jaiswal, Rahul Maini and Mohan Pedhapati.
They had already spent years studying how computer systems can fail.
Jaiswal and Maini had previously discovered a serious problem in Apple’s systems.
Apple paid them $50,000 for reporting that problem.
The researchers said they used AI to make the difficult exploit-writing work faster.
The important skills, such as choosing where to look and how to connect weaknesses, still came from the people.
They reported the OpenAI issue responsibly and did not harm the company.
Harsh Jaiswal, Rahul Maini and Mohan Pedhapati co-founded Hacktron AI in 2025.
The Indian-origin researchers used AI assistance to help develop a reliable exploit against OpenAI systems.
Jaiswal and Maini previously found an Apple infrastructure misconfiguration that enabled access to protected files and code execution.
Apple paid the researchers $50,000, while OpenAI paid $6,500 for the later disclosure.
The researchers worked through OpenAI’s official program, reported the findings and caused no harm.
- Who
- Indian-origin security researchers Harsh Jaiswal, Rahul Maini and Mohan Pedhapati, co-founders of Hacktron AI.
- What
- They used AI assistance to develop an exploit that reached OpenAI code and reported it through OpenAI’s official program.
- Where
- The work involved OpenAI systems and previously Apple’s infrastructure; the article does not specify the researchers’ location.
- When
- Hacktron AI was founded in 2025; the researchers’ Apple work occurred around January 2021.
- Why
- They were conducting white-hat security research and reporting vulnerabilities responsibly.
AI Lowers the Barrier
Expertise Remains Essential
Meaning of the breach
AI Lowers the Barrier
The incident could be read as evidence that AI has made sophisticated hacking accessible to almost anyone.
Expertise Remains Essential
The researchers’ track record indicates that experienced security professionals supplied the key knowledge and judgment.
AI’s contribution
AI Lowers the Barrier
AI helped accelerate the most time-consuming part, writing a reliable exploit.
Expertise Remains Essential
AI compressed the labor but did not decide where to look, which vulnerabilities to connect or when to stop.
Key facts
- Researchers
- Harsh Jaiswal, Rahul Maini and Mohan Pedhapati
- Company founded
- Hacktron AI was co-founded by the researchers in 2025.
- AI mentioned
- Claude was used to assist the OpenAI research.
- OpenAI payment
- OpenAI paid $6,500 for the disclosure.
- Apple payment
- Apple paid $50,000 for an earlier disclosure.
- Earlier finding
- Jaiswal and Maini found an Apple misconfiguration that enabled access to protected files and code execution.
- Research conduct
- The researchers used OpenAI’s official program, reported everything and caused no harm.
Sources
Related news
Researchers Used Claude to Reach OpenAI Internal Systems
Researchers Reveal OpenAI Agents Probed Hugging Face Before July Breach
OpenAI Code Access Highlights Falling Costs of AI Security Attacks
AI Agents Allegedly Breached Hugging Face and Tried Cover-Up
Russian-Speaking Hackers Used Cursor AI Against Seven Companies
Two AI Security Incidents Put OpenAI Under Scrutiny
Lawsuit Claims AI Giants Coordinated to Slow Development
Anthropic’s Annualised Revenue Tops $100 Billion Ahead of IPO
