Editorial illustration for ChatGPT Crafts Childbirth Course, Saves Hours of Research
ChatGPT Creates Childbirth Course in Hours
ChatGPT Crafts Childbirth Course, Saves Hours of Research
Three people and a chatbot got into OpenAI's private codebase faster than most companies can patch a printer. That's the headline out of Hacktron AI's disclosure this week, and it lands in the same stretch of July when OpenAI itself admitted its models had broken into Hugging Face. The company that builds the tools now finds itself on the other end of them.
Hacktron's team didn't need a nation-state budget or months of reconnaissance. They used Anthropic's Claude to help write the attack, chaining an image-upload flaw on OpenAI's community forum into a way of hijacking employee accounts. The whole thing took under 72 hours. OpenAI paid out $6,500 for the bounty, which is a modest sum given what the researchers say they could reach.
The bigger story is what this says about AI-assisted hacking right now. Hacktron claims the same underlying flaw let it into Slack, Meta, and GitHub too, suggesting this wasn't a one-off bug so much as a pattern showing up across major platforms. Here's how the researchers describe getting in.
Security startup Hacktron says its three-person team reached the AI giant's private code in less than three days, then reported the hole and collected $6,500 for it.
Why this matters
We've got two versions of AI trust playing out in the same week, and they don't square easily. On one end, a parent uses ChatGPT to build a labor course and cheat sheet in an afternoon, work that would've taken hours of scattered googling and forum-diving. That's the pitch working as intended: real time saved, real utility, no asterisks.
On the other end, a three-person outfit walks into OpenAI's private code in under three days using Claude as the lockpick, and gets paid $6,500 for the privilege of telling them about it. Hacktron played nice. The next team through that door might not.
For builders and founders leaning on these models daily, the lesson isn't "AI is good" or "AI is risky," it's that both are true at once and neither cancels the other out. The same systems generating genuinely useful personal tools are running on infrastructure that a small, motivated team can crack in 72 hours. If you're shipping products on top of OpenAI's stack, that bounty payout is worth more attention than the next mini-course headline.
Related Reading
Browse all OpenAI news →
Open Source D&B rebuilds 642 million‑business database after AI agents hit limits
Open Source Box CEO Aaron Levie Says AI Will Drive "Biggest Cybersecurity Upgrade
Open Source Study: 6.7% of Deepfake Requests on Hugging Face Targeted Children
Related Topic OpenAI researcher quits, citing distrust over ad‑driven engagement metrics
LLMs & Generative AI
Related Topic OpenAI launches GPT-Image 1.5 with precise editing for enterprise visuals
LLMs & Generative AI
From Archives Amazon Raises Privacy Concerns Over Meta's Stealth AI Agent
From Archives Runway and Google DeepMind Pursue Real-Time, Interactive AI Video Generation
From Archives Anthropic Delays IPO After OpenAI Despite Growing Enterprise Revenue
Business & Startups
From Archives Unity Launches Official Plugins for Claude Code and OpenAI Codex
Business & Startups
Common Questions Answered
How did Hacktron's three-person team access OpenAI's private codebase so quickly?
Hacktron used Anthropic's Claude to help write the attack and chain together exploits to breach OpenAI's private code in less than three days. The security startup then responsibly reported the vulnerability and received a $6,500 bug bounty from OpenAI for their discovery.
What is the significance of OpenAI being hacked by a small team using Claude?
This incident demonstrates that sophisticated security breaches don't require nation-state budgets or months of reconnaissance when AI tools like Claude can be leveraged to automate attack development. It highlights a paradox where the same AI companies building security tools are themselves vulnerable to attacks using competing AI models.
How does the ChatGPT childbirth course example contrast with the security breach mentioned in the article?
The childbirth course represents the positive use case of ChatGPT saving parents hours of research and forum-diving by generating educational content in an afternoon. In contrast, the security breach shows the darker side where AI tools can be weaponized to compromise even major AI companies' infrastructure within days.
What does the article suggest about AI trust and security in the same week?
The article presents two conflicting narratives about AI trust: one showing genuine utility and time-saving benefits for consumers, while the other reveals that major AI companies like OpenAI remain vulnerable to sophisticated attacks. These simultaneous events highlight the difficulty in reconciling AI's beneficial applications with its security risks.
Further Reading
- How Hackers Used Claude Opus 5 to Breach OpenAI Systems - Security Boulevard
- This tiny cybersecurity startup managed to hack OpenAI using Claude - Business Insider
- Anthropic's Claude helped cybersecurity researchers breach OpenAI - Anadolu Agency
- OpenAI hack: How 3 Indian-origin researchers used Anthropic’s Claude to access employee accounts - The Indian Express
- Three researchers hacked OpenAI using Claude Opus 5 to do it - daily.dev
