3 guys hacked OpenAI using a rival Anthropic model. Here's what to know.
Summary
Three cybersecurity experts in India used an advanced AI model from the company Anthropic to hack OpenAI's community forum accounts. They showed how powerful AI tools can make it easier to find security weaknesses and access private user information, including that of OpenAI employees.Key Facts
- Three researchers from the Indian company Hacktron used Anthropic’s AI model Claude to hack OpenAI in July.
- They targeted a security weakness in a third-party tool called Discourse, which OpenAI uses for its forum.
- Using the more advanced Claude Opus 5, they accessed some OpenAI users’ ChatGPT and Codex accounts.
- The hack could have given access to linked apps like email and Slack, exposing private conversations.
- The team was conducting this as part of a "bug bounty" program, which rewards finding software flaws legally.
- Without AI help, the hack would have taken months, but with Claude it took less than three days.
- Hacktron researchers have also hacked other big tech firms like Apple, Google, and Microsoft in the past.
- The researchers warned that AI’s rapid progress increases risks and companies must improve their security.
Read the Full Article
This is a fact-based summary from The Actual News. Click below to read the complete story directly from the original source.
