Malicious ChatGPT Share Links Could Hijack Sessions to Run Code and Read Private Chats

What happened

Researchers at the security firm Parameter found a flaw in ChatGPT's share-link feature, according to reporting by The Hacker News. A shared link crafted by an attacker could run the attacker's code in the logged-in session of anyone who opened it. That access could let the attacker read the victim's private conversations.

Parameter reported the issue to OpenAI through its bug bounty program on October 18. OpenAI shipped a fix eight days later and rated the flaw P1, its most serious category.

Why it matters

Share links are a common way people use ChatGPT. Users send and open links from colleagues, team chat channels, and documents. A flaw that turns opening a shared link into code running inside the victim's logged-in session is a session-hijack risk. It needs only a click, not stolen credentials. Private chats are a concern because sensitive business and personal data often flows through these conversations.

The specific technical root cause has not been detailed publicly. The available reporting does not describe evidence of exploitation before the fix shipped.

What defenders should watch for now

• Treat ChatGPT and similar AI-platform share links as untrusted content. Apply the same caution used for other link-based attacks.
• If your environment logs web traffic in detail, review browser and endpoint logs for unusual activity following visits to ChatGPT share-link URLs.
• OpenAI has patched this server-side, so there is no local detection rule to deploy. The main actions are user awareness and confirming your AI usage policy covers this risk.
• Watch for similar share-link risks in other AI chat platforms. The underlying issue is not unique to OpenAI.

Developing story

This is a vendor-disclosed and patched vulnerability with limited public technical detail so far. The available reporting does not mention a CVE. We will update if more detail emerges. Read the original report at The Hacker News.